Performing FIB on Extremely Rough Surfaces

FIB is one of the most powerful techniques for characterizing thin films. Often when we need to see below the surface of a device, FIB is the most reliable and versatile technique. When paired with SEM imaging and elemental analysis, it is capable of thoroughly characterizing the dimensions and composition of entire device stacks. Often we use this technique to cut into a flat surface such as a metal pad or a passivation layer in order to see the structure that lies beneath. However, there are circumstances where FIB must be performed on a much less ideal surface. For example, surfaces with extremely high roughness can make ion milling much more complicated. Protruding features can block the cut from view due to the tilted electron beam imaging angle, and dramatic ridges and valleys can lead to ion beam scattering, in some cases making it extremely difficult to perform cuts in certain regions. It is common in some industries to create such a surface; for example, some dental implants need extremely high surface roughness in order to adhere to their target area. An example of an etched composite implant surface is shown below[1]:

Challenges

It is sometimes the case that we have to analyze a sidewall profile for such a surface. For example, a material such as a sintered metal may have a thin coating applied to it. In order to reveal the coating sidewall profile using FIB, especially if the RMS roughness is in the micron range, it is often necessary to be very strategic about cut type and location, and it may be necessary to apply a thick platinum coating first. Additionally, the lower current and multiple passes required to get a high level of cut precision may not be worth the time investment. Optimizing FIB process flow in order to achieve high quality results in time is extremely important, especially in fast-paced environments where cross sectional analysis must be performed multiple times a day. Finally, this article discusses an alternative solution, which involves making a very large, deep cut at high current in order to bypass some of these issues.

The liquid metal ion source (LMIS) uses relatively heavy ions accelerated at a high voltage, such as 30kV. These ions are generally heavy enough to mill most materials, including hard composites and metals. However, certain undesired effects such as curtaining and uneven milling occur when scanning the ion column over a region that contains heterogeneous materials and structures. A high-roughness surface can magnify this effect to the extreme, resulting in ion mills that sometimes are entirely unable to reveal a decent sidewall profile, even if a cleaning cut is performed. In this case, certain approaches to cut optimization can help mitigate these effects. 

Understanding Milling Angle

Depending on the incident angle of the ion beam on the surface, there are significant variations in the milling rate. At normal incidence (where the ion beam is perpendicular to the surface), the milling rate is relatively low. This is because the momentum transferred from the ions to the atoms in the lattice is primarily directed into the material, rather than toward ejecting atoms from the surface. As a result, the energy is dissipated deeper into the lattice, which can cause subsurface damage without efficiently removing material from the surface.

In contrast, when the ion beam is tilted at an angle relative to the surface, the momentum transfer is more effective at ejecting atoms from the surface. This is because the component of the momentum parallel to the surface helps atoms overcome the binding energy of the lattice and escape more easily. Consequently, a tilted incidence angle results in a higher milling rate, as more surface atoms are sputtered away with each ion impact. The figure below shows how normal incidence (a) results in a much lower ejection rate than an angled incidence (b). [2] 

While there is variation of the angle resulting in peak sputter yield between materials, the sputter yield is highest at 75-80 degrees for many common materials, including silicon. For example, the image below shows sputtering yield versus incident angle for cobalt and tungsten carbide, which follow a similar curve to silicon. [3]

Because of this effect, protruding structures that are flat on top can be difficult to cut evenly using FIB. An example of this is shown below. At normal incidence, the top surface of this pyramid structure mills more slowly than the sides due to incidence angle, resulting in a protruding structure, even after significant milling. [4] Thus, it is important to consider this effect when milling an extremely rough surface. Milling a large area with significant topography can leave behind large pillars and protrusions that have flat surfaces on top.

Curtaining and Redeposition Effects

In addition to beam angle considerations, other effects like curtaining and redeposition can result in rough, low-quality sidewall profiles. For example, the image below shows curtaining (a) and redeposition (b) for a large FIB cut of a porous surface. Depositing a metal layer over the region of interest can help mitigate these effects and is discussed in the following section. [5]

Additionally, due to milling angle effects, the curtaining phenomenon can be partially mitigated by “stage rocking”, which applies small periodic variations to the stage angle during the mill. Newer systems such as the FEI Helios 5 PFIB include stage rocking features; however, most FIB systems don’t have this feature, and other methods must be employed to minimize the curtaining effect. [6]

Addressing Uneven Milling

One approach to milling an uneven sample would be to use lower beam current, more passes, and a smaller, more surgical cut that is optimized over a small set of topography. For example, removing the front portion of a tall ridge with carefully tuned beam parameters could ensure that the tilted SEM imaging is not obstructed by prominent features. And cutting a small region off of a targeted ridge could allow for a cleaner side wall. 

However, there are significant challenges involved with this approach. First of all, lowering the beam current and using multiple passes takes more time. A typical FIB cut at a higher current might take a couple of minutes, while a highly surgical cut could take tens of minutes or longer. In an environment where it’s important to optimize cost and rapid results, this analysis flow may not be practical.

Additionally, for ridges with high curvature, the mill rate can vary significantly due to the changing angle of incidence across the surface as discussed previously. This variation complicates the milling process, as geometric factors must be carefully accounted for to achieve consistent results. Successfully performing such precise milling on highly rough or irregular samples demands considerable expertise and experience, further increasing the complexity and time required.

Depositing a Metal Layer

There are also other considerations to take into account depending on the depth of a film or other feature to be analyzed. Will the film be damaged by the ion or electron beam? Is there charging, drift, or other beam artifacts? If so, it may be necessary to deposit a layer of metal such as platinum over the area to be analyzed. This can also help mitigate some of the more dramatic effects of surface roughness on the evenness of the milled region. However, a thin coating of metal may not be enough to neutralize these effects, and significant curtaining and cut irregularities may persist. Furthermore, depositing metal, especially using the electron beam, adds significant time to the analysis flow.

One approach would be to deposit a much thicker Pt layer, on the scale of microns, using a high current I-beam deposition. This could ensure a much cleaner cut can be performed. It may be necessary to first perform a thin e-beam deposition in a selected region, perhaps no more than 300nm, in order to protect the surface during high current deposition. Then, a higher current deposition can be performed, up to a few microns of material if necessary. For example, a sample with 2um surface roughness and maximum feature variation over 10um may need a 3-5 um deposition in order to achieve a smooth region of metal that fully covers the surface. While this deposition adds time to the analysis flow, it can be used to get a much cleaner sidewall profile without dramatic curtaining and other irregularities. A wide cross section can be cut, revealing the sidewall profile of a much larger region of the sample that would be allowed with a surgical cut. An example process flow is shown below, using a GIS needle to deposit platinum:

1. Perform 300nm E-beam Pt deposition over region of interest

2. Tilt and perform medium-current I-beam Pt deposition, approximately 1um thick

3. Perform final high-current Pt deposition, up to 3um if necessary

4. Perform high current FIB cut over a wide area, scanning upwards towards the middle of the deposited region.

5. Perform lower current FIB cleaning cut on the revealed sidewall

Such a process flow for FIB would allow any surface coatings to be protected, while also being faster than more surgical approaches such as low-current, mult-scan cuts. And a large region of the sidewall could be imaged this way, allowing for the observation of the larger scale topography of a coating or structure.

However, there are still challenges with this method. First of all, the large scale and current of this cut could damage the surface and lead to significant redeposition. Samples that are sensitive or need to be subsequently re-imaged in the same area may not be a good fit for this method. And large scale deposition may not fully cover the valleys of the sample, leading to voids between the platinum layer and the sample surface at the bottom of these valleys. It is important to make sure that these voids are noted as analysis process artifacts rather than part of the device profile. It may be necessary to selectively image certain regions of the exposed sidewall in order to avoid showing these voids in any final results.

FIB remains one of the most powerful and versatile techniques for nanoscale device analysis, despite challenges that may arise from certain surface features. The deposition of a metal layer may be necessary to handle highly irregular surfaces, protect films, and prevent artifacts like charging and drift. It is important to consider total analysis time, comprehensiveness of imaging, and sensitivity of surface features when performing such analysis. By understanding the tradeoffs and drawbacks involved with performing cross-sectional imaging on samples with extremely high surface roughness, it is possible to optimize the analysis flow for a wide range of requirements.