Nanofabrication

Positive and negative resists

In a positive resist, the exposed regions become more soluble in the developer and are removed. In a negative resist, exposure causes the written regions to become less soluble, so the exposed material remains.

The choice depends on the application rather than one resist type being universally better. Important considerations include achievable resolution, sensitivity, resist thickness, etch resistance, pattern density, lift-off geometry, process compatibility, and writing time.

Dose and process window

Exposure dose determines how much electron energy is delivered to the resist. Too little dose can leave features incompletely developed or undersized, while excessive dose can broaden features, close small gaps, or increase the influence of proximity effects.

In practice, an EBL process is usually developed through dose testing rather than relying on a single nominal value. The goal is not simply to find one dose that works, but to identify a useful process window where the required dimensions remain stable despite small variations in exposure or development.

Proximity effects

Electrons do not deposit all of their energy exactly where the beam enters the resist. Forward scattering broadens the beam as electrons travel through the resist, while electrons that penetrate the substrate can backscatter and expose surrounding regions.

As a result, dense and isolated structures can receive different effective doses even when the nominal writing dose is identical. Proximity effects can alter line width, spacing, corner definition, and the uniformity of repeated features.

This becomes particularly important when a pattern contains regions with very different local densities. Dose correction, geometry adjustment, resist choice, substrate effects, and development conditions may all become part of the optimization.

Development is part of the lithography process

Exposure changes the chemistry of the resist, but development determines which part of that modified resist actually remains on the sample. Developer composition, time, temperature, agitation, rinse conditions, and resist thickness can therefore change the final critical dimension and profile.

I find it more useful to treat exposure and development as one coupled process rather than optimizing them independently.

Contact and proximity lithography

In contact lithography, the photomask is brought into direct contact with the resist-coated substrate. Minimizing the gap helps preserve image fidelity because diffraction has less distance over which to spread the projected pattern.

In proximity lithography, a small separation is intentionally maintained between the mask and wafer. This reduces mechanical contact and the risk of mask or wafer damage, but the additional gap increases diffraction and can reduce resolution.

Exposure dose and focus

Exposure dose must be high enough to create the required chemical change through the resist thickness, but excessive exposure can broaden features or reduce the separation between nearby structures.

Focus and mask-to-wafer spacing are closely related to image quality. Small deviations that are insignificant for large features can become important as the critical dimension approaches the optical resolution limit.

This is why dose matrices and controlled process tests are often more useful than assuming that a single exposure setting will work equally well for every resist thickness, substrate, and feature geometry.

Resist thickness matters

Resist thickness is a process-design choice rather than simply a coating parameter. A thicker resist may provide better protection during a later etch or create more vertical space for lift-off, but it can also make small features more difficult to resolve.

The required thickness therefore depends on what the resist must do next: survive a plasma, define a lift-off profile, protect an underlying film, or simply transfer a relatively shallow pattern.

Direct-write optical lithography

Maskless direct-write systems expose the resist according to a digital layout rather than using a physical photomask. They occupy a useful space between conventional photolithography and EBL.

They generally do not provide the ultimate resolution of EBL, but they can dramatically simplify design iteration because a new pattern does not require fabrication of a new photomask.

This makes direct-write lithography particularly useful for prototyping, larger device features, contact levels, test structures, and process development where the design may change frequently.

Mask selectivity

Selectivity describes how quickly the target material is removed relative to the mask. If the mask etches too quickly, it can lose thickness before the desired etch depth is reached.

Mask erosion can also change the critical dimension during the process. For deeper etches or more aggressive plasma chemistries, the mask may need to be thicker, more resistant, or replaced by a hard mask such as a dielectric or metal layer.

Anisotropy and sidewall profile

Many device structures require material to be removed primarily in the vertical direction. An anisotropic etch preserves lateral dimensions better than an isotropic process and can produce steeper sidewalls.

But verticality alone does not guarantee a good profile. Bowing, undercut, footing, taper, roughness, redeposition, and sidewall damage can all affect the usefulness of the final structure.

In practice, sidewall shape is often just as important as etch depth because it can influence electrical isolation, optical behavior, subsequent conformal coating, and metallization.

Chemistry versus physical bombardment

Dry etching often relies on a combination of chemical reactions and energetic ion bombardment. The chemical component provides material selectivity through volatile reaction products, while ion bombardment helps drive directional removal and can assist reactions at the surface.

Increasing physical bombardment can improve anisotropy, but it can also increase surface damage, mask erosion, and redeposition. Increasing the chemical contribution may improve selectivity, but excessive lateral chemical etching can reduce profile control.

Process development therefore becomes a balance between etch rate, anisotropy, selectivity, surface condition, and dimensional control.

Loading and pattern density

Etch behavior can depend on how much exposed material is present in the chamber. Large open areas and dense arrays of small features may consume reactive species differently, creating local or wafer-scale differences in etch rate.

This means a recipe that performs well on a small test pattern may behave differently when transferred to a denser design or a larger patterned area.

Chamber condition and process history

Plasma tools are not perfectly independent of their usage history. Residues from previous materials, chamber wall condition, cleaning, seasoning, and contamination can influence plasma chemistry and reproducibility.

When an established process suddenly changes, chamber history is one of the variables worth checking before immediately modifying the recipe.

ICP power: controlling plasma density

In an ICP system, the inductively coupled source generates a high-density plasma. Increasing ICP power generally increases the population of ions, electrons, radicals, and other reactive species available to participate in the etch.

A higher plasma density can increase etch rate, but more plasma is not automatically better. Changes in species concentration can alter selectivity, surface chemistry, mask behavior, and sidewall passivation.

ICP power therefore needs to be considered together with bias power, pressure, gas chemistry, and the material being etched.

RF bias: controlling ion energy

The substrate bias influences the energy with which ions reach the wafer. Increasing ion energy can enhance directional material removal and help produce more anisotropic profiles.

However, excessive ion bombardment can increase mask erosion, surface damage, sputtering, redeposition, and loss of selectivity. For sensitive materials or interfaces, minimizing unnecessary ion damage may be just as important as achieving the desired etch depth.

One advantage of ICP-RIE is therefore the ability to maintain a relatively dense plasma while independently adjusting the ion energy reaching the substrate.

Chamber pressure and directionality

Pressure affects how frequently particles collide while travelling through the plasma sheath and chamber. At lower pressure, the mean free path is generally longer, allowing ions to reach the substrate with fewer collisions and better directional control.

At higher pressure, more collisions can broaden the distribution of particle trajectories and modify the balance between chemical and physical etching.

Pressure also influences plasma stability, residence time, reactive species concentration, and transport of reaction products, so its effect cannot be considered independently from gas flow and power.

Gas chemistry depends on the material

The choice of plasma chemistry depends strongly on the material being etched and the volatility of the products that can be formed. Fluorine-based chemistries are widely used for many silicon-containing and dielectric materials, while chlorine-based chemistries are commonly used for a range of III-V semiconductor and metal-containing systems.

Additional gases can be introduced to modify sidewall passivation, dilution, ion bombardment, polymer formation, or the concentration of reactive species.

The relevant question is therefore not simply whether a chemistry can etch a material, but whether it can do so with the required selectivity, profile, surface quality, and compatibility with the rest of the device.

Selectivity and mask survival

Selectivity compares the removal rate of the target material with the removal rate of another material, often the lithography mask or an underlying layer.

If the mask selectivity is insufficient, the resist or hard mask can erode before the required depth is reached. Mask erosion can also change the lateral dimensions of the feature during the etch.

For this reason, resist thickness and mask material should be selected together with the expected etch depth and process chemistry rather than treated as independent decisions.

Profile control is often the real objective

Etch rate alone provides very little information about whether a process is suitable for device fabrication. Two recipes can remove the same film at similar rates while producing completely different sidewall profiles.

Undercut, taper, bowing, roughness, footing, mask erosion, and redeposition can all affect subsequent processing and ultimately device performance.

This is why cross-sectional or high-resolution inspection can be much more informative during process development than simply measuring how much material was removed.

Resist profile and undercut

An undercut profile is often helpful for lift-off because the resist sidewall recedes beneath the top surface. This reduces the chance that deposited material forms a continuous bridge from the substrate to the top of the resist.

Bilayer resist stacks are commonly used when a controlled undercut is needed. The lower resist layer can develop laterally beneath the upper imaging layer, creating a geometry that favors discontinuous deposition.

The required undercut depends on the film thickness, deposition method, feature size, and resist thickness. Too little undercut may not solve bridging, while excessive undercut can weaken the resist structure or distort small features.

Film thickness versus resist thickness

Resist thickness should be chosen with the final deposited film in mind. If the deposited layer becomes too thick relative to the resist profile, material can coat the sidewall and connect the top film to the film on the substrate.

A thicker resist can provide more vertical separation, but it can also reduce lithographic resolution or make small features harder to define. As with many fabrication choices, the useful solution is a balance rather than simply maximizing one parameter.

Directional and conformal deposition behave differently

Evaporation is relatively directional because atoms or molecules travel predominantly along line-of-sight trajectories from the source toward the substrate. This makes e-beam and thermal evaporation particularly compatible with lift-off because sidewall coverage can remain limited.

Sputtering is generally less directional. Scattering in the process gas and the broader angular distribution of arriving species can improve sidewall and step coverage, which is valuable for many applications but can make lift-off more difficult.

The same property can therefore be beneficial or problematic depending on the fabrication goal: conformality is useful when continuous coverage is required, while directionality is often preferred when clean lift-off is the priority.

Step coverage and device continuity

A directional film can produce excellent lift-off but may not cover vertical or sloped device topography well. Conversely, a more conformal film can improve electrical continuity across steps while becoming harder to pattern by lift-off.

This becomes especially important when contacts need to reach across dielectric edges, etched mesas, or other three-dimensional structures. The deposition method should therefore be chosen based on the complete device geometry rather than only the lithography process.

Adhesion and interface preparation

Even a geometrically perfect metal pattern can fail if the film does not adhere well to the underlying surface. Native oxides, organic residue, moisture, contamination, or incompatible materials can weaken the interface.

Surface preparation and adhesion layers are therefore often part of the metallization process. The appropriate approach depends on the substrate, contact material, device function, and whether the interface is intended to be electrically active.

Lift-off is also a cleaning problem

Once deposition is complete, the resist still needs to be removed without damaging the desired structure. Solvent choice, soaking time, temperature, agitation, and mechanical assistance can all affect the result.

Aggressive ultrasonication may remove stubborn material, but it can also damage delicate features. For small or fragile structures, patience and controlled solvent access can be more important than mechanical force.

Process windows matter more than perfect recipes

A recipe that produces an excellent result at one exact condition is not necessarily a robust manufacturing process. Real fabrication contains variation: film thickness changes slightly, resist coatings vary, chambers drift, substrates differ, and measurements contain uncertainty.

A useful process therefore needs a window in which reasonable variation still produces an acceptable result. Understanding that window also helps identify which parameters are truly critical and which have more tolerance.

For me, process development is therefore not simply about finding the best-looking sample. It is about finding conditions that can be repeated reliably.

Geometry can change the process

A process developed on one geometry does not always behave identically when feature size, pitch, aspect ratio, or pattern density changes. Deposition and etching are three-dimensional processes, so the available space around a structure can influence how material reaches or leaves the surface.

For example, a deposition that fills a relatively open structure may behave differently in a narrow gap, while a conformal film can gradually close an opening from the sidewalls before the underlying volume has been completely filled.

This is why scaling a device is not always equivalent to scaling a drawing. Changes in geometry can require changes in the fabrication strategy itself.

Metrology closes the process loop

Process integration depends on knowing what actually happened at each critical stage. SEM, profilometry, ellipsometry, reflectometry, AFM, XRD, electrical measurements, and cross-sectional analysis can provide different pieces of that information.

I think of metrology as part of fabrication rather than something that happens after fabrication. Measurements create the feedback loop needed to compare the intended process with the structure that was actually produced.

When a device fails, that process history can be used to work backward from the final symptom and identify where the structure first departed from the expected result.

Troubleshooting means separating symptoms from causes

The step where a failure becomes visible is not necessarily the step that caused it. Poor electrical performance after metallization, for example, could originate from the metal itself, but it could also reflect an earlier surface-cleaning problem, incomplete etching, dielectric residue, or damage introduced upstream.

I therefore prefer to troubleshoot by building hypotheses from the process history and available measurements, then testing those hypotheses systematically rather than immediately changing multiple recipe parameters.

Root-cause analysis starts with evidence

When a previously stable process moves outside its expected window, I first look for what changed. That can include the incoming material, resist or chemical lot, tool condition, chamber history, recipe revision, operator sequence, upstream processing, or even the measurement method used to evaluate the result.

Run records, travelers, tool logs, metrology, and comparison with successful baseline samples help narrow the possibilities. Once a likely cause is identified, controlled experiments can distinguish correlation from the actual mechanism.

Change one thing, understand everything it touches

Process changes are sometimes necessary because of tool availability, scaling, yield, material changes, or a newly identified failure mode. But even a seemingly small modification can affect several downstream operations.

Before adopting a change, I find it useful to define what problem the change is intended to solve, what new risks it introduces, how success will be measured, and which downstream steps need to be revalidated.

This makes process development more deliberate and helps prevent solving one problem by quietly creating another.