Combinatorial thin-film screening, as a service
We make the material you want to try.
Including the compositions nobody sells, will not make one-off, and will not run as a gradient. One wafer carries 342 of them, measured on the same grid, in a week.
- 342 compositions per wafer
- Dozens of elements deposited
- 100 mm wafer
All 37 elements are modeled here. Nitrogen and oxygen can also be added from gas.
Scroll
What we can make
Pick two to seven elements. Nitrogen and oxygen can also be added from gas, so oxides and nitrides are possible without needing them as a separate solid source.
What you can order
Three things, each ordered on its own. A screen is not required for the other two.
One targeted deposition
The composition is already decided and you need the film made. One deposition, to your specification, no screen and no gradient.
One composition, off gradient
A composition you have already chosen, deposited uniform, the same all the way across.
A full library screen
342 compositions in one run on one grid, for when the composition is the thing you are still looking for.
Why 342 and not twenty
Twenty samples spread across a ternary sit far enough apart that an entire region, or a phase boundary, can fall between two of them and never be seen.
Why measured rather than calculated
Simulation returns the energy of a structure you propose. Published growth models give a shortlist. Neither tells you which structure formed on your substrate, in your chamber, at your conditions. One run answers that for 342 compositions at once.
How a run works
-
01
Deposit
Every source fires at once, and the balance between them sets the gradient. One wafer comes out carrying 342 different compositions.
-
02
Map
Composition, structure and the property you came for, measured at every position, and every instrument works off the same grid, so any two measurements of a position are measurements of the same place.
-
03
Choose
The map shows where the property peaks and where it collapses. That one composition is deposited again as a uniform film, the same all the way across.
Run one composition at a time and the same search is measured in months or years.
What we deposit
Combined onto one substrate, with the power method matched to each element. A material that resists steady bombardment, an insulator, a magnetic metal, a refractory metal, still gets a matched delivery instead of one setting for everything.
Process
Substrate temperature and the electrical bias applied to it are both set per run. Nitrogen or oxygen can be introduced during a run, so oxides and nitrides form directly instead of needing a separate step.
Elements
Dozens of elements are available, common ones already in stock and others ordered in as needed. Nitrogen and oxygen can be added during a run to form oxides or nitrides. Purity is specified per element.
Substrates
100 mm wafers, and larger substrates for some processes. Semiconductor, oxide and single-crystal substrates, or bring your own.
Formats
Composition libraries of 342 graded compositions, uniform films off-gradient, layer stacks, test layers on your substrate, or your component itself.
Anneal
Vacuum anneal and rapid thermal anneal, temperature and hold time both set per run.
Material classes
Piezoelectric and radio frequency (RF) films. Hard and wear coatings. Transparent conductors and other functional oxides. Contacts, barriers and metallizations. Superconducting films. Refractory and high-entropy solid solutions. Rare-earth-free magnetic films. Shape memory alloys. Thermoelectrics. Battery electrode and solid electrolyte films. Electrocatalysts.
If your class is not on this list, ask anyway. We take on systems we have not run before, and we will say plainly whether yours is workable.
Cross-contamination
A few elements contaminate other runs in the same system and need a dedicated one; say early if one of those is unavoidable.
Within a run the process is held fixed and the composition is graded, so every position on the wafer grew under the same conditions. We choose those conditions and we state them with the map. The same library can be carried into a second state: measured as deposited, annealed, measured again, one wafer and two states.
Against one composition at a time
The conventional route
The material has to already exist, made in advance at your exact composition and purity. If it does not, someone has to fabricate it first, before any trial can start. Either way, it is committed before the search is known to be worth running.
One composition in one process state per run, each measured on its own sample, so no two maps land on the same place.
Months, sometimes years.
One run at xemx
We combine pure elements directly during deposition, so a composition can be tried without a pre-made alloy existing first. We keep the common elements on hand and order in the rest.
342 compositions in one process state on one grid, so the maps lay over each other; a second run gives the same library in a second state.
A week for most metal and alloy systems.
What we measure
On your run, composition, structure and the target property are measured and mapped at all 342 positions. The wafer here is calculated, like every graphic on this page; scroll to see what each measurement returns, drawn on it. The supporting methods run on the positions the map has already picked out, which is where they are worth their time.
Elemental compositionAll elements, 342 positions
Drag to turn the wafer.
Illustrative examples, by property
Composition: a smooth gradient, by design
Each source lays down more of itself close by and less further off, so composition shifts gradually across the wafer. It is the one quantity here that does move smoothly. That is what makes it the baseline: where structure steps or a property peaks, it does so against a composition that did not.
Structure: a step at a phase boundary
Composition changes smoothly across the wafer, made in one run. Crystal structure often does not: where one phase gives way to another, it happens as a sharp step, not a gradual shift, because the two phases are different structures.
Electrical: a resistivity peak, not a straight line
Sheet resistance can rise toward the middle of a composition range and fall again near either pure end. A mixed lattice scatters electrons more than a pure one does, most of all near equal parts of two elements.
Mechanical: hardness split between composition and grain size
Hardness tends to rise with solute content, because strain in the lattice blocks dislocations from moving. Growth conditions also set grain size across the wafer, and a finer grain raises hardness on its own. The two effects can add up or work against each other from spot to spot.
Thickness: a geometric fall-off from each source
Material arriving from a fixed point falls off with distance and angle, in a pattern set by that point's own geometry. A wafer fed by several fixed points tends to show thickness highest near each one and falling off toward the outer edge, an effect of geometry alone, apart from any electrical or mechanical property.
Electrochemical: a threshold, not a gradient
Activity can change sharply rather than smoothly with composition. Many alloys stay active until a key element reaches a critical amount and builds a stable oxide layer on the surface. Past that point, activity can drop in a step rather than a smooth trend.
Microstructure: a shift in growth mode
Surface roughness depends on how a film starts to grow, and that growth mode can shift with composition and local arrival rate across a wafer. A change from layer-by-layer growth to island growth can move roughness more than either mode's own smooth trend does. That is why spot checks sit where composition or structure has already flagged a change.
Scroll to map each property on the wafer, one at a time.
Composition
Elemental composition at every one of the 342 positions. Oxidation state on selected areas. Depth profiles, light elements, hydrogen content, and trace elements are available on request. Ions can be implanted into a library before it is measured.
Structure
Phase, texture, lattice parameter, crystallite size, and strain. Every position, one scan.
Thickness
Thickness across the whole wafer. It varies along with composition, so it is mapped like the rest rather than quoted as one number. Density is derived from mass per unit area and thickness, not measured.
Mechanical
Hardness and modulus at 342 areas. Stress and thermal expansion, from separate reference structures, not this map.
Electrical and optical
Sheet resistance at 342 areas. Optical and magneto-optic response. Sheet resistance and magnetic response are also measured across a temperature range, because a film that has to survive heat is not captured by a room temperature number.
Electrochemical
Electrochemical activity across the library, under light where the response is photoelectrochemical. Finer spacing where the map needs it.
Microstructure
Morphology and roughness, as spot checks on positions the map has already picked out.
What you get back
The map
Composition, structure and the target property tied to each of 342 positions, with the candidate and the ruled-out regions marked.
The wafer
The measured library itself, one 100 mm wafer holding 342 compositions. It can come to you; say so and we will arrange it.
The film
One composition off the map, deposited again as a uniform film, the same all the way across.
Where the method comes from
Combinatorial screening grew up inside university groups and corporate R&D departments, and almost none of them run projects for anyone else. That is the whole reason this is hard to buy. xemx runs these projects for outside clients.
xemx is a Ruhr-Universität Bochum spin-off, built on the library method from Alfred Ludwig's group and the electrochemical screening method from Wolfgang Schuhmann's group. Both are published: Ludwig, npj Computational Materials 5, 70 (2019) on the library method, and Sliozberg et al., ChemSusChem 8, 1270 (2015) on the electrochemical screening method.
Where to start
Tell us what you are working on, or the material you are curious about and cannot easily try, in whatever terms you have. Nothing needs to be settled before you write. The details belong on a call.
If you are not sure this process can reach your material at all, ask anyway. We will say plainly whether it is workable, and we can carry a result across into your own process.