Characterization of Surface Modifications by White Light Interferometry: Applications in Ion Sputtering, Laser Ablation, and Tribology Experiments

Example 1

WLI is not widely used for surface characterization in tribological work, but it is in fact a powerful method for quantitative measurement of wear volumes for many contact geometries. WLI produces a full 3D representation of the surface that can be analyzed using any of several visualization software packages. These packages enable various types of measurements to be performed. For greater lateral resolution, images can be “stitched” together to produce wide-area information (several mm), with μm resolution.

For non-tribological work, WLI can be used to measure surface features which are difficult to measure using AFM or other contact means. Fu et al. studied the effect of Ga ion beam parameters on the micromachining of Si surfaces. AFM was used to measure the surface profile, but the results were limited due to the restricted vertical range of the AFM cantilever tip, and damage caused by the tip for large pit depths. Instead, the workers found WLI to be more useful for measuring large depths while preserving features and being able to easily measure vertical sidewalls.¹⁷

In tribology, analysis software can extract statistical surface roughness parameters of worn surfaces, which can be compared with analytical models of surface topography generated by tribological processes, e.g. machining. Jiang et al. used this capability to study surface roughness as a function of milling parameters and compare with analytical models.¹⁸

A very simple use of WLI is for measuring the material loss from a wear scar during unidirectional sliding of a pin on a flat disk. Reiter et al. used WLI to scan wear tracks and calculate the wear volumes and hence wear rates for a series of ball on disk tests. Because the 2D surface is reconstructed, it is a simple matter to determine the depth and breadth of a wear scar, and thus calculate wear volume. In this sense, WLI serves as nothing more than an enhanced stylus profilometer.¹⁹

More powerfully, WLI can be used to measure the volume of wear lost by a sliding contact if the profile of the original surface is known. In the simplest example, a groove or crater is worn into a flat surface. The wear volume is simply the volume of material that is removed from the flat surface. Surface analysis programs enable the volumetric deviation from a flat surface, i.e. the volume lost, to be measured. These measurements are easiest when the original counterface is a symmetrical surface: flat, spherical, or cylindrical. Devillez et al. sketches out a method by which WLI was used to measure the volume lost from the surface of a cutting tool. The local area of the flank was initially flat, and it was relatively simple to calculate the difference between the original surface and the wear groove that was produced on the surface.²⁰ When the surface is not flat, then an additional step in the procedure is necessary in order to measure wear volume. If the original surface is geometrically regular, then it is possible to mathematically remove the curvature of the surface so that a flat plane is produced, while simultaneously deforming the wear scar in the same manner. The deviation from the plane can easily be calculated.

In biomedical tribology, WLI, sometimes called vertical scanning interferometry in the medical literature, can be used to measure the surfaces of native and worn articular cartilage. Topographic information can be obtained, but the technique is hindered by the fact that the surface of live cartilage is dynamic and moves as water is lost or absorbed.²¹

With this example we tried to show how WLI can be used for routine analysis in tribology; additional background can be found in Refs. 22, 23, and some references therein.

Example 2

In many experiments designed to determine sputtering yields (SY) of various materials under specific ion bombardment conditions, uncertainties in ion beam parameters can propagate and result in uncertain sputtering yield values.²⁴ For example, it can be challenging to determine shapes of ion beam profiles and the corresponding operational current densities, especially when the projectile energy goes below 1 keV and then further approaches the sputtering threshold. Moreover, under such conditions, the focusing of the ion beam is in question, and the relative spread Δε/ε in the initial kinetic energy distribution of ions²⁵ can have strong influence on experimental results.^{9, 26}

By combining the WLI visualization with precise measurements of the total ion current by a Faraday cup (FC), the SY and the operating current density can be obtained simultaneously. Besides, this approach appears to be very helpful in estimating the extent of undesirable “wings” of the ion beam profile to help in aligning ion beam sources. The sputtering yield Y is then estimated by the following expression

where I, direct current of an ion beam; τ, time of sputtering; M_atom, mass of a matrix atom in grams; ρ, density; e, the elementary charge. V is the volume of the removed sample material obtained by means of the WLI measurement. Volume calculations can be performed either by using both approaches described in the Protocol for flat surfaces, depending on a type of post-processing software available, or by three dimensional integration based on cross sections in two orthogonal directions centered on the eroded surface area (black lines in Figure 8a and 8b) simply by

in the event that one has no option to perform detailed post-processing as described in the Protocol. The |Max-Min|_Yprofile parameter in the second multiplier term is used not to take the crater depth into account two different times, when calculating the V.

The ion beam currents are measured in situ by a custom graphite FC consisting of an internal pin (inlet holes of 250 μm dia.) and the external surface. This design provides coarse control over the ion beam focusing conditions by measuring “internal” and “external” (mostly attributed to “wings”) components of the delivered current. Position of the FC fully simulates a sample surface positioning with respect to focusing and directing ion optics of our mass spectrometer.²⁷ The measurement of energy spread Δε of the low energy system²⁷ can be accomplished using the same FC. In this case, FC can be externally biased by a voltage 0 to 5 kV, and the total current as a function of suppression voltage is measured. In such a manner, the inherent Δε was estimated to be 23 eV.

The symmetric profile that is seen in Figure 8 suggests that there is good alignment of the ion beam column, and FWHM of 120 μm at a total current of 2 μA. The WLI approach allows one to characterize the ion sputtering with the identical normally incident ion beam decelerated to 150 eV by the target potential. In this case, the cross section of the static beam spot is shown by a blue dotted line, and the crater cross section is shown by cyan open diamonds. The ion column enabled delivery of the same 2 μA of Ar⁺ current on the target because the deceleration of the beam from the nominal 5 keV energy to 150 eV occurred in the immediate vicinity of the target, and in such a way that its optimal focusing was maintained by an electrostatic lens (the FWHM of 150 μm in Figure 8c is evidence of that).²⁷ The sputtered crater has in this case a larger lateral size because the deflection voltages of the raster-generating octupole were kept unchanged for the two primary ion impact energies, resulting in additional beam swinging due to the target potential.

Based on the WLI data, sputtering yields of Cu(110) at 5 keV and 150 eV ion impact energies were determined. An obtained SY value of 1.8 at/ion for the former case was in good agreement with literature data.²⁸ For the latter one, the sputtering yield was 0.2 at/ion (rare or lack in the literature).

For SY estimates, our results demonstrate an alternative experimental approach, which can also be used to verify experimental data and to “tweak” adjustable parameters of predictive models^{29, 30} and calculation codes such as SRIM³¹ and TRIDYN,³² as well as to generate reference data for many technological applications.^{6, 9, 33-36} In addition, this approach is capable of accurate quantifying sputtering rates for organic materials and solids under bombardment with a variety of primary species, commonly used atomic ions and relatively new molecular and cluster ions, such as in Refs. 37, 38. Thus it helps to resolve a problem of time (or primary ion fluence) to depth conversion in depth profiling experiments by using an average parameter called the sputtering rate where d is a total depth measured by WLI and τ, as mentioned above, total time of sputtering.

Example 3

As in the case with ion beam sputtering, the ablation yield is an important parameter for analytical applications. For ablation, this value is usually expressed in terms of removal rate per shot, or alternatively removal rate per unit of time with a given laser repetition rate. Because there is minimal thermal loading of materials, repetition rates may be very high (MHz) and is often limited by the rate at which the beam can be moved over the material. Additionally, there are multiple material ablation thresholds, corresponding to different fundamental ablation mechanisms.³⁹ The predominant analytical tools using ultrafast laser ablation require high fluences (>5 J/cm²) and correspondingly high removal rates and consume relatively large quantities of sample for analysis.^{40, 41}

In principle the ions formed during the ablation process can be analyzed directly, or ablated neutrals can be ionized with a second laser, leading to a more sensitive and higher spatial resolution technique. As is seen in Figure 9a, two craters ablated under identical conditions, with the exception of fluence, will have dramatically different shapes. These craters are representative of a larger set of craters formed on a polished single crystal GaAs sample. By simply decreasing the fluence from 1 J/cm² to 0.4 J/cm² (the latter value being about twice the ablation threshold for GaAs),⁴² the crater diameter is reduced by nearly half, and the average removal rate at the center of the crater is reduced from about 10 nm/shot to 5 nm/shot. The mass removal rate is apparently reduced by a factor of six, improving the corresponding analytical volume resolution.

It is important to note that the pronounced ring structure observed around each hole (Figure 9a) should be interpreted with care, since it may be due to scattering of the light from the interferometer. Although it is possible to get ring-shaped crater material ejecta, they are generally only seen with much longer pulsed lasers.²⁷ This optical artifact can be expected if the radius of the crater’s edge is near, or below the diffraction limit for the average wavelength in the far field conditions of the microscope (roughly 1 μm in this case). In such situations, if measurement of the crater wall radius is crucial a complimentary technique such as those described above should be employed – see the following section on artifacts and limitations of WLI. However, if the main goal is to measure the crater depth, with a secondary goal to ensure that a threshold radius near the diffraction limit is not exceeded, then WLI is well suited for quickly measuring large numbers of craters.

Some typical artifacts/errors and limitations of WLI

1. Error estimation and uncertainty assessment

All real surfaces are rough and irregular to some extent, and no measurement technique is perfect. These imperfections will introduce uncertainties in the measurement of material loss. There are three sources of error. First, there is measurement error and noise inherent in the profilometer. Second, errors may arise due to imperfect curvature removal if this step is performed. Third, the original undamaged surface may be rough and irregular, which can lead to subtraction errors because the original surface is not known to perfect accuracy. An estimate of the magnitude of the total error can be obtained by simply performing several measurements on pristine sample surfaces using the identical processing/levelling technique as done on a modified sample surface. Figure 14 shows one of five measurements made of pristine ball surfaces. The resulting average volume was 92 μm³ with a standard deviation of 184 μm³, instead of the expected value of zero. This indicates that for these samples and specific processing technique used, there is statistical variation of 184 μm³ and a systematic error of 92 μm³.

Also, let us briefly describe typical artifacts and limitations one meets when using WLI approach. These artifacts contribute additional uncertainties which affect precision of final results. For each specific WLI application instance, they should be addressed separately.

2. A single or multilayer constituent film of optical properties that differ from those of a base/substrate. Transparent/semitransparent film on a reflective base

The simplest type of WLI requires optical specular reflection from a surface. If the entire surface has the same reflectivity in terms of phase change of the reflected wave, then a measurement accurate to nm will be obtained. For inhomogenous samples (for instance, integrated circuit patterns), a correction needs to be applied based on the nature of the particular material. This effect is addressed in a paper by Harasaki et al. ⁴³ Offsets as large as 36 nm may occur for Ag/Au pairs. In such cases, detection part of WLI can be easily “confused” by different optical responses in terms of phase shifts in a way that gives a wrong estimation of the full depth. It is also common, if one sees that a bowl is inverted into a pedestal that sticks above the zero level of a pristine surface; the height is skewed as well. This effect is nicely demonstrated in Figure 13 in Ref. 7. A straightforward way to avoid such problems is to uniformly cover the entire surface with some known reflective material so it will have the same optical response to a WLI detector. Then this measurement problem disappears. In Ref. 7, this problem was resolved through covering of a layered system, SiO₂ on top of Ti base, with gold film. At the same time, it is worth mentioning that in some cases one might try to deal with optically transparent layers. The basis of such a measurement by WLI is somewhat different and has to do with the limited coherence length of natural (white) light. It was demonstrated that transparent films of ≈10 nm and higher thickness on a reflective substrate can be measured by WLI. A good description of this procedure can be found in Refs. 44 and 45.

3. Batwings

As mentioned with respect to what is shown in Figure 9 above, sharp features such as steps and crater edges may cause diffraction of the light from the interferometer, leading to anomalous signals known as “batwings.”⁴⁶ In general, this occurs when the lateral dimensions of such features are near the diffraction limit and the step height is below the coherence length of the light. Experimental approaches have been developed to minimize such artifacts (see for example Ref. 47). Figure 15 demonstrates this effect in case of ion sputtering experiments. The green solid line is a crater produced by a “sharply” focused Ar⁺ ion beam of ≈30 μm dia. that gave abrupt change in heights between the surface and the crater’s bottom with vertical walls. On the one hand, it suggests a very good ion beam alignment, but at the same time those artifacts called “batwings” are introduced by WLI into the final profiles reconstruction, and so should be excluded from volume removed calculations. There is a dotted cyan line in the same plot, which represents exactly the same sputtering conditions but under the ≈60 μm dia. beam. As one can clearly see, “batwings” completely disappeared. This is due to pronounced change in transition gradient as the focusing condition differs.

Concluding Remarks

It is important to keep in mind if a research work connected with a new type of sample is started using WLI it is always a good idea to establish whether or not WLI is suitable for those purposes. If yes, then it is necessary to calibrate/check calibration using an independent approach; and only after that WLI becomes an express method of getting bulk amounts of results. There are three of them as was already mentioned: AFM, SEM and stylus. Figures 9b and 16 are examples of comparing of WLI results to SEM and stylus, respectively.

Figure 9b answers a question if the outer ring spikes of ejected material of larger ablated craters seen in WLI pictures are real. It is hard to determine precisely, but results of performed SEM imaging compare favorably with what WLI provides in terms of height of sharp ring structure observed around each hole ≈400 nm (≈500 nm from WLI profile) and ≈12 μm dia. (≈13 average dia. deduced from WLI profiles).

Figure 16 is an example of how WLI profiling results can be confirmed by stylus profiler, if dimensions of modified area are suitable for this purpose. The idea behind this experiment was to check depth calibration through an independent technique. The object is a crater obtained by Ar⁺ 5 keV ions in Si substrate, the lateral dimensions, lying beyond AFM capabilities, are such that use of stylus is favored. In Figure 16 two profiles obtained through WLI and stylus superimposed one onto another. Direct comparison of the data suggested that results obtained by WLI in terms of depth are correct. The lateral dimension is also reproducible with an only exception: laterally crater measured by stylus looks narrower against WLI. Taking that the sample is single-component and all the transition spatial gradients are small, it is reasonable to assume that WLI data reflect actual size of the removed area, and smaller size in stylus representation is due to convolution of the real crater size with characteristic probing tip size. As a rule, depth calibration of our WLI profilometer is accomplished through Ted Pella AFM step-like standard of 500 nm height.