FYNE – a next generation X-ray source for high-tech industries

1a. Fast measurement at nanoscale resolution - background

Understanding the interplay between resolution and dose in nanoscale imaging is critical to ensuring an efficient imaging setup. There are two main effects that compete: electron trajectory and electron penetration. We will briefly explain how this competition arises and what strategies exist to deal with it. In an X-ray tube, electrons are ejected from an emitter and accelerated under high voltage in vacuum onto a target material. The resulting collision of electrons with the target atoms produces X-rays [3]. The 3D interaction region within the material determines the focal spot volume of the X-ray source. At higher acceleration voltages the trajectories of electrons originating from the emitter are less able to spread due to the influence of the stronger electric field between the anode and cathode.

The result is a smaller lateral spread of the electrons reaching the target (see Fig. 5a), hence a reduction of the focal spot size and an increase of the resolution. However, at higher acceleration voltages, the electrons also have higher kinetic energy as they reach the target. Higher kinetic energies result in an increase of the average time between collisions with the atoms of the target material [4]. This means electrons can propagate further into the target, inducing additional lateral spreading through scattering with the target atoms. As a result, the focal spot volume increases and the ultimate resolution decreases (see Fig. 5a). The balance of these two effects determines the focal spot size.

This makes it clear that the target thickness plays an important role in determining the X-ray interaction volume and therefore the focal spot size. In a thick target the electrons can scatter laterally into the volume of the target, and this scattering increases at higher acceleration voltages. The increased volume results in a larger focal spot which limits the resolution, but also provides higher dose for the same reason. In a thicker target, the resolution is limited primarily by the scattering of electrons that can occur throughout the 3D volume.

One way to achieve higher resolution is therefore to limit the thickness of the target (see Fig. 5B). This reduces the potential for lateral propagation of the electrons within the material and limits the focal spot volume. The disadvantage to this approach is quite clear: reducing the interaction volume reduces the maximum X-ray dose that can be produced. 

1b. Fast measurement at nanoscale resolution- example 

Let’s now take two examples to illustrate these effects in the real FYNE 160 NF4 source. In Fig. 6 we show data for two possible source configurations across the accessible voltage range (60 – 160 kV). In the first example, using a thicker target, the spatial resolution is approximately constant across the entire voltage range at around 1.2 µm. However, the dose reaching the detector varies by almost 80% over the same range. In addition, compared to the dose in the second example using a thinner target the dose achieved in this case is higher at all voltage points, and considerably higher at larger voltages. The disadvantage to this arrangement is that the dose varies strongly at different operating points, and the ultimate resolution is limited. The higher dose and limited resolution can be understood from the discussion above: a larger interaction volume produces more dose, but also increases the lateral focal spot size. These settings are therefore appropriate for applications requiring consistent resolution across a large voltage range and higher X-ray flux. 

The second configuration shown in Fig. 6 uses a thinner target than the first. In this case, the behavior across the voltage range is completely different to the thicker target. In particular, the resolution changes at different acceleration voltages. 

This is due to the stronger electric field resulting in a smaller spread on the target, which is not cancelled out by lateral spreading within the target. The target chosen here is designed to maximize dose while achieving a resolution of 0.4 µm (400 nm) at a specific operating point of 160 kV. However, as is clear in the comparison with the first case, the total dose obtained is less than for the thicker target. In general, higher spatial resolution means reduced X-ray flux. Therefore, when planning nanoscale imaging it is critical to consider the real operating point of the source and the resolution required for an application, and to maximize the dose for these requirements. Selecting for the highest achievable resolution will mean reduced dose, and if such a high resolution is not required will result in a significant loss of inspection efficiency.

The key takeaway is therefore: it is critical to optimize the source for the specific requirements of an application - FYNE allows this flexibility.

2. Exceptional focal spot stability

Given the precision required to perform sub-micron imaging measurements, the X-ray source also needs to be highly stable to prevent the “virtual” focal spot size becoming too large for the application. The virtual focal spot is the size of the focal spot plus any movement due to drifts averaged over the duration of a measurement (see Fig. 7). Typical measurement lengths may be between a few seconds for 2D images to many hours for detailed 3D scans. Therefore, low drift behavior can be critical. 

To illustrate the high stability of the FYNE source, we performed drift measurements over a period of 24 hours. No adjustments or corrections were made to the source during that time. The source, sample manipulator and detector are connected via a granite block within the X-ray cabinet to minimize the influence of vibrations. To determine the drift, we did the following. A sample with a circular gold structure was placed in the center of the field of view and imaged at a fixed operating point of 120 kV. The position at time T is determined by subtracting the image of the gold structure taken at time t = T from the image of the same structure taken at t = 0. The pixel shift is then converted into a real space shift in X and Y directions. The measurements are performed adjustment-free across the entire time range i.e. the beam position is not controlled in any way during the entire measurement. 

Important to note is that it is not possible to fully disentangle a shift of the image caused by intrinsic drift of the focal spot from a drift caused by changes in the rest of the imaging chain e.g. thermalization of the mechanical supports or thermalization of the sample (which is placed directly in front of the source). Initially there is a period during which the entire mechanical system thermalizes. 

The measurements are summarized in Fig. 7. The result is the combined stability of the entire imaging chain, not only of the focal spot. Following thermalization of the system, the drift velocity in the Y-direction is 0.029 µm/hour, while in the X-direction it is only 0.012 µm/hour is observed. The maximum total deviation of the focal spot across the measurement is therefore below the resolution of the source (0.4 µm). These values are up to a factor 100 smaller compared with some alternative open tube designs. Such a small drift velocity is remarkable given that no active beam stabilization is employed at any point during the measurement.

3. Long source lifetime

Since FYNE is an “open” X-ray source, meaning it can be physically opened and maintained allowing replacement of critical parts, the lifetime of the source is technically unlimited. This is a factor to consider when calculating the total cost of ownership of an inspection system, as sealed tubes must be replaced at regular intervals. Nonetheless, in inspection applications it is crucial to maximize uptime and reduce the regularity with which maintenance must be performed, especially if used in a production environment for at-line or even in-line processes. An especially sensitive component in nanoscale imaging applications is the electron emitter unit, which in classical open tube design has a lifetime on the order of hundreds of hours, although this varies widely depending on the operating parameters. As an example, assuming an average operation of only 8 hours per day, a filament with 500 hours of lifetime would need to be exchanged every 2 months. Extending this period is critical to increasing uptime in both lab and industrial settings. 

We performed a lifetime measurement using the imaging setup of Fig. 2. The metric chosen for the measurement was the spatial resolution in the X and Y directions over time. A 0.6 µm structure was imaged at an operating point of 120 kV. The FYNE source was run continuously for the duration of the test, without any pause in operation. The results are summarized in Fig. 8. No significant change in resolution is observed even after more than 2600 hours of source operation. Under the previous example of 8 hours per day operation, this would translate to around 11 months of operation without maintenance. Certainly, many factors play a role in determining the source lifetime and the parameters that are chosen as a metric are application specific and cannot be generalized here. However, the measurements highlight the order of magnitude of source lifetime that can be achieved with FYNE: thousands of uninterrupted operating hours.