Resolving Mass Spectral Overlaps in Atom Probe
Tomography by Isotopic Substitutions: Case of
TiSi15N
David Engberg, Lars J. S. Johnson, Jens Jensen, Mattias Thuvander and Lars Hultman
The self-archived postprint version of this journal article is available at Linköping University Institutional Repository (DiVA):
http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-122721
N.B.: When citing this work, cite the original publication.
Engberg, D., Johnson, L. J. S., Jensen, J., Thuvander, M., Hultman, L., (2018), Resolving Mass Spectral Overlaps in Atom Probe Tomography by Isotopic Substitutions: Case of TiSi15N,
Ultramicroscopy, 184, 51-60. https://doi.org/10.1016/j.ultramic.2017.08.004
Original publication available at:
https://doi.org/10.1016/j.ultramic.2017.08.004
Copyright: Elsevier
Resolving Mass Spectral Overlaps in Atom Probe
Tomography by Isotopic Substitutions â Case of TiSi
15N
David L. J. Engberg1, Lars J. S. Johnson2, Jens Jensen1, Mattias Thuvander3, and
Lars Hultman1
1 Department of Physics, Chemistry and Biology (IFM), Linköping University,
SE-581 83 Linköping, Sweden
2 Sandvik Coromant, LerkrogsvÀgen 19, SE-126 80 Stockholm, Sweden 3 Department of Physics, Chalmers University of Technology,
SE-412 96 Göteborg, Sweden
Abstract
Mass spectral overlaps in atom probe tomography (APT) analyses of complex compounds typically limit the identification of elements and microstructural analysis of a material. This study concerns the TiSiN system, chosen because of severe mass-to-charge-state ratio overlaps of the 14N+ and 28Si2+ peaks as well as the 14N2+Rand 28Si+ peaks. By substituting 14N
with 15N, mass spectrum peaks generated by ions composed of one or more N atoms will be
shifted toward higher mass-to-charge-state ratios, thereby enabling the separation of N from the predominant Si isotope. We thus resolve thermodynamically driven Si segregation on the nanometer scale in cubic phase Ti1-xSix15N thin films for Si contents 0.08 †x †0.19 by APT,
as corroborated by transmission electron microscopy. The APT analysis yields a composition determination that is in good agreement with energy dispersive X-ray spectroscopy and elastic recoil detection analyses. Additionally, a method for determining good voxel sizes for visualizing small-scale fluctuations is presented and demonstrated for the TiSiN system.
Keywords: atom probe tomography (APT); mass spectral overlap; isotopic substitution; isotope enrichment; titanium silicon nitride (TiSiN); time-of-flight mass spectrometry (TOFMS)
1. Introduction
Atom probe tomography (APT) is gaining grounds in many fields of materials research. The combination of 3D analysis, sub-nanometer resolution, and elemental identification makes it a powerful complement to crystal structure sensitive techniques, such as X-ray diffraction (XRD) and transmission electron microscopy (TEM). The sample of interest is shaped into a sharp tip, up to ~200 nm in apex diameter, cooled to cryogenic temperatures and kept in ultra-high vacuum [1]. Tip apex atoms are field evaporated [2] by either field or thermal pulses, possibly undergoing post-ionization to higher charge states [3], and subsequently accelerated toward a detector. Ideally, no more than one ion should be generated per pulse. Measurement of the time between field evaporation and detection enables elemental
identification of the ions by their mass-to-charge-state ratio. The position-sensitive detector, combined with sequential field evaporation, makes 3D reconstruction of the tip possible by back-projection of the detected ions [4]. Traditionally, field evaporation was induced by a combination of a DC voltage and short voltage pulses. Conductive materials were a
prerequisite, since the field induced stress would otherwise prematurely destroy the
samples [5,6]. By replacing voltage pulses with laser induced thermal pulses [7,8], the span of available materials systems has been broadened to include semiconductors and insulators, as well as brittle conductors [9,10].
In addition to the challenges in developing application modes of APT, the new materials systems in themselves increase the complexity. Hard ceramic materials can be used as an example to illustrate this, as field penetration is deeper and bonds are generally stronger in ceramics, as compared to metals, due to their covalent nature [1,11,12]. Covalently bonded materials are more likely to generate molecular ions in an evaporation event than metals [13]. Single and molecular ions combined with several charge states of each ion lead to mass
spectra that often have closely spaced peaks, and possibly complete overlaps, even in materials systems with only a few elements, e.g., binary FeCr used in stainless steel
production or ternary ceramics like TiSiN [11]. Peak decomposition techniques can be used to estimate the relative amount of each element that two or more overlapping peaks consist of, but correct positions of the corresponding ions cannot be ascertained. Some overlapping ions, in particular ions with different charge states, could be separated by kinetic energy analysis, but this would require detector techniques that are not yet available for commercial
instruments [14].
TiSiN is a materials system where severe overlaps impede the analysis of APT data. Ever since the first reports were published on superhard TiSiN nanocomposite films grown by plasma enhanced chemical vapor deposition (PE-CVD) [15], the field has attracted the
attention of both materials scientists and industry. The commercial success of TiSiN has made it a subject of a vast number of analysis techniques other than APT [16â19]. Currently, thin films of TiSiN made by cathodic arc deposition are used extensively in cutting tool
applications to increase the lifetime of the tools. While there is no thermodynamically stable ternary phase of TiSiN [20], it can be considered pseudo-binary, consisting of cubic TiN and SiNy phases for Ti:Si ratios up to at least x = 0.2 (while PE-CVD typically yields an amorphous
grown by cathodic arc deposition to be a metastable cubic phase mixture rather than a nanocomposite of the thermodynamically stable phases TiN and Si3N4 [19,21].
Outstanding questions for such TiSiN films concern the Si segregation, structure, phase composition, and bonding in as-synthesized or annealed TiSiN, on a finer length scale than obtained by analytical and high resolution TEM. Furthermore, increased Si content and
extended annealing promote the formation of Si3N4 or SiNy tissue phases of crystalline [22] or
amorphous [21,23] nature covering the TiN grains. Here, the local nitrogen stoichiometry is unknown, as is the retained Si fraction in the TiN phase. Such data is needed in a functional sense for tailoring the microstructure of the film and thus hardening, as well as eventually controlling mechanical deformation and wear mechanisms.
The composition sensitivity and high spatial resolution of APT make it an excellent analytical instrument for the task outlined above. However, due to the severe mass spectral overlaps of the 14N and 28Si peaks, few attempts to study TiSiN using APT have hitherto been
made, and these yielded inconclusive results [21,24]. The peaks of 14N+ and 28Si2+ are only
separated by 0.0145 Da, making them practically impossible to separate with commercially available atom probes at present, even though they can be partly resolved under favorable conditions [25]. However, by replacing 14N with 15N, most overlaps in the mass spectrum can
be avoided without making any significant change to the chemistry of the material. A similar study on FeN enriched in 15N by Sha et al. [26], and more recent studies of isotopic
enrichment in SiO2 and SiN using 18O and 15N by Kinno et al. [27,28], have been conducted. In
this paper, we show that the most abundant isotopes of Si and N can be separated in the APT mass spectrum using TiSi15N thin solid films produced by cathodic arc deposition. The
usefulness of isotopic substitutions is demonstrated by comparing differences in simulated mass spectra from natural and isotope-enriched systems, and by evaluating a typical
experimental mass spectrum from 15N-enriched TiSiN. General requirements regarding which
types of materials systems that are suited for applying isotopic substitutions are discussed, and a number of such systems are identified.
2. Experimental details
Samples for the APT mass overlap separation study were TiSi15N thin solid films grown by
reactive cathodic arc deposition in an industrial system (Sulzer Metaplas MZR323) described below.Atom probe tips were prepared from TiSi15N films using the protocol described in
detail by Thompson et al. [29]. An approximately 15x2x4 ”m3 strip of substrate and film was
removed from the coated WC-Co plate using the lift-out technique in a Zeiss 1540EsB CrossBeamÂź focused ion beam (FIB) station. The end of the strip was mounted by Pt
deposition on top of a truncated Si post, the mounted part of the strip was cut off and the process was repeated for the entire length of the strip. This procedure was done for films of two different Si concentrations. Each mounted sample was subsequently shaped into a sharp tip using annular milling. The acceleration voltage was decreased from 30 kV to 5 kV in the finishing step to reduce implantation of Ga+ ions.
The tips were analyzed by an Imago LEAPÂź 3000X HR in laser pulsing mode. The pulse
frequency was 200 kHz while the pulse energy was either 0.4 nJ or 0.6 nJ. Even though Tang et al. [30] reported that mass resolution is improved with pulse energies above 1.0 nJ, the pulse energy in this study was kept low because an initial parameter study on the similar TiN system showed significant loss of N at pulse energies of 0.9 nJ and higher. The sample temperature was set to 60 K in all but one analysis, in which 30 K was tested. The pulse energy and temperature were varied between analyses to validate that the DC voltage was low enough not to cause field evaporation between pulses. Such evaporation would increase the background noise and, as TiSiN is a multi-component material, possibly distort the measured composition. The DC voltage was regulated to maintain a specific evaporation rate of the order of 0.002 to 0.005 detected ions/pulse. Reconstructions and mass spectrum simulations were made using Cameca IVAS©. Since no crystallographic features could be
identified in APT, the reconstructions were made using tip profiles from scanning electron microscopy (SEM) micrographs acquired in the FIB before the APT analysis.
The generalized lambda distribution was chosen to describe the peak shape of the simulated mass spectrum. When parameterized by Freimer et al. [31], the inverse of that distribution is given by Eq. 1, in which đąđą varies linearly from 0 to 1, đđ1 is the transversal shift,
đđ2 is the scale factor while đđ3 and đđ4 determine the peak shape. The values of đđ1 and đđ2 were
manually varied between peaks while đđ3 and đđ4 were manually set to 0.2 and -1.3,
respectively, to mimic the peak shapes of experimental TiSi15N spectra.
đčđčâ1(đąđą) = đđ 1+đđ1 2ïżœ đąđąđđ3 â 1 đđ3 â (1 â đąđą)đđ4â 1 đđ4 ïżœ (1)
The system for growing the TiSiN films used four 63 mm cathodes with different Ti:Si atomic ratios that were mounted vertically (100:0, 90:10, 80:20, and 75:25 from top to bottom) on the inside of the chamber door, while three 100 mm pure Ti cathodes were mounted
vertically on the chamber wall. The substrates consisted of polished 12x12x4 mm3 plates of
cemented carbide (WC-Co), except for the samples designated for time-of-flight energy elastic recoil detection analysis (ToF-E ERDA). These were cut into pieces of approximately
12x6x1 mm3 prior to the deposition. The samples were mounted in the deposition system
using magnets in two vertical rows on opposite sides of a rotating drum. Each row consisted of seven positions at different heights. Four positions were at the same height as the Ti:Si cathodes, with one additional position between each. The positioning of cathodes and samples created a compositional gradient between the films. During all depositions the substrate temperature was set to 450 °C, the gas pressure of natN2 or 15N2 was 2 Pa and the bias voltage
was -30 V. The substrates were sputter cleaned and a diffusion barrier layer of TinatN was
deposited on all samples using the rotating drum and the three 100 mm cathodes. The
rotation of the drum was then stopped and the first row of substrates was positioned directly in front of the Ti:Si cathodes before being coated with TiSinatN. Without breaking vacuum, the
gas line was filled with 15N2 and the chamber was evacuated from natN2. At the same time, the
drum was rotated 180° in order to position the second row of substrates in front of the Ti:Si cathodes. These substrates were then coated with TiSi15N.
Plan-view samples for TEM investigation were prepared by mechanical polishing and
subsequent Ar-ion polishing in a Gatan Precision Ion Polishing System. TEM micrographs and electron diffractograms were recorded with a Philips CM20 ST TEM instrument at an
acceleration voltage of 200 kV. Energy dispersive x-ray spectroscopy (EDS) and ERDA were used to measure the elemental composition of the films. The EDS analysis was conducted in a Leo Gemini 1550 SEM instrument equipped with an Oxford Instruments EDS, using an
acceleration voltage of 20 kV. The ERDA were performed using the set-up at Uppsala University [32]. A 36 MeV 127I8+ primary ion beam was directed onto the samples at an
incident angle of 67.5° with respect to the surface normal and the induced recoil ions were detected at an angle of 45° with respect to the incoming ion beam. The measured ERDA spectra were converted into relative atomic concentration profiles using the CONTES code [33].
X-ray diffractograms were recorded using a Panalytical Xâpert MPD Bragg-Brentano diffractometer with Cu-Kα radiation by attenuation of Cu-KÎČ with dual Ni filters. The
diffractometer was optimized for fast measurement by using scanning line mode with the Xâcelerator multiple strip detector. A 10 mm brass mask and a 0.5° divergence slit were used to limit the irradiated area and collimate the beam, while the primary and secondary
anti-scatter slit was 0.5° and 5 mm, respectively.
3. Results & Discussion
3.1 Film microstructure
XRD measurements in the Ξ-2Ξ configuration and the range 35° †2Ξ †65° of all as-deposited Ti1-xSixN films are presented in Fig. 1. The diffractograms show that the films are single-phase
cubic with mixed grain orientations for x †0.11. With higher Si content, the texture becomes (002) and the diffraction peak FWHM increases, which indicates a decrease in average grain size and possibly an increase in microstrain from lattice defects. Any remaining (111) signal can be attributed to the TiN diffusion barrier.
The stable positions of the (002) peaks in Fig. 1 reveal that the lattice parameter of these Ti1-xSix15N alloys is constant around 0.424 nm regardless of Si content, which also is the
nominal value for TiN. This is in agreement with our studies on the system with TiSiN solid solution thin films grown using Ti-ion assisted arc deposition [19] or high-power pulsed magnetron sputtering [34]. Optical inspection of the film corresponding to x = 0.15 reveals a speckled pattern, whereas the other films are even in coloration, changing from golden yellow to brown copper with increasing Si content. SEM micrographs of the same film (not shown) reveal differences in coating thickness, which causes the intensity of the (002) peak to be lower than expected for x = 0.15.
Fig. 1. XRD diffractograms from Ti1-xSix15N thin film samples with Si contents of
0.01 †x †0.19 as determined by ERDA (Section 3.4). Nominal 2Ξ values for TiN (111), (002), and (022) are marked by dotted lines, while substrate peaks from the WC-Co are marked with
stars (*).
3.2 APT Mass spectra
The range of interest in the mass spectrum of one typical Ti0.81Si0.1915N atom probe analysis is
presented in Fig. 2 (a). Peaks of particular interest, such as 15N+, 15N2+, 15N22+, 28Si+, 28Si2+, and 28Si15N22+, as well as the isotopic signature of Ti2+, Ti3+, and Ti15N2+, have been tagged. The Ti
and TiN isotopic signatures consist of five visible peaks each, where the center peak is significantly larger than the surrounding four. The peak at 15 Da contains an unavoidable overlap of 15N+ and 15N22+. Normally, the magnitude of each component in this overlap could
be estimated by comparing the 14N15N2+ peak at 14.5 Da in the N2 isotopic signature. In this
case, however, any 14N15N2+ signal would be lost in the much larger 29Si2+ peak due to the
scarcity of 14N in the sample. Since this prevents peak decomposition, it is not possible to say
with certainty how much the different ions contribute to the 15 Da peak. Important clues are the insignificant 15N2+ (0.06% area compared to the 15N+ peak in Fig. 2 (a)) and unresolvable 15N23+ peaks at 7.5 Da and 10 Da, respectively (not shown). Based on these indications,
combined with the metastable nature of 15N22+ [35], the 15 Da peak was ranged as 15N+ ions
only. Kinno et al. [28] and Sha et al. [26] reached the conclusion that the 15 Da peak in their experiments consisted mostly of 15N+ by investigating samples with different 14N:15N ratios.
As the natural abundance of 15N is very low, peak decomposition when using natN would be
challenging at best, making this overlap an intrinsic problem in the analysis of nitrides. Because the extent of the improvement when substituting natN with 15N cannot be easily
assessed from Fig. 2 (a) alone, we consider the simulated mass spectra of TiSi15N and TiSinatN
presented in Fig. 2 (b) and Fig. 2 (c), respectively. Relevant parts of a mass spectrum from an experimental APT analysis of nanocomposite TiSiN with similar composition as ours
(~8 at. % vs. 9.8 at. % Si) and N15 < 0.4% (the natural abundance of N15 being 0.364 % [36]),
can be found as âSample 2â of Fig. 2 in the paper by Tang et al. [24]. A comparison of this mass spectrum with that in Fig. 2 (c) shows a high degree of similarity in peak positions and
relative sizes. The simulated peaks rise rapidly to the left of the peak positions and then decreases exponentially to the right, and while the peaks in both Fig. 2 (a) and the mass spectrum by Tang et al. [24] follow this trend, their rise is not as steep. The mass spectrum by Tang et al. [24] shows small but distinct C2+ and N2+ peaks at 6 Da and 7 Da, respectively.
These were not included in the simulations because the corresponding peaks in the
experimental mass spectrum are negligible in area. Clearly, the C content is higher and the charge state ratio for N (and possibly C) is different in the analysis by Tang et al. [24]
compared to this study. As can be seen in Fig. 2, the overlapping peaks at 14 Da and 28 Da in (c) are well separated in (a) and (b), now located at 14 Da and 15 Da as well as 28 Da, 29 Da and 30 Da. Unfortunately, the 30Si2+ peak is located at 15 Da, while the 30Si+ and 30Si15N22+
peaks are located at 30 Da.
Fig. 2. Partial mass spectra from (a) an APT analysis of a Ti0.81Si0.1915N sample as well as mass
spectrum simulations of (b) Ti0.81Si0.1915N, and (c) Ti0.81Si0.19natN made using Cameca IVAS©.
These overlaps cannot be avoided without also substituting natSi with 28Si, but as the natural
abundance of 28Si is 92.2 at. % and that of 30Si is 3.1 at. %, the percentage of resolved Si ions is
still significantly increased compared to samples grown with natN. The 46Ti14N2+ peak at 30 Da
is shifted to 30.5 Da when using 15N, thereby avoiding a complete overlap with 30Si. Partial
46Ti15N2+ at 30.5 Da will occur, however, only at ~2% of the 15 Da and 30 Da maximum peak
height. Furthermore, there is a discrepancy in the isotopic signature of Ti15N2+ compared to
Ti2+ and Ti3+, barely visible in Fig. 2 (a), where the peak at 31 Da is slightly larger than
expected. This could be explained by the fact that Si is a strong hydride former [14]. Some 30Si
may have been misinterpreted for TiN and if so, some 29Si must have been mistaken for N2
and some 28Si for 28Si15N22+. Nevertheless, since both 29Si and 30Si are much less abundant than 28Si, the isotopic substitution should significantly improve the mass spectrum and allow most
Si to be distinguished from N and vice versa. To verify this, the composition measured by APT must be worked out carefully in concert with other techniques, as covered in Section 3.4.
3.3 Field evaporation behavior in APT
The mass spectrum in Fig. 2 (a) shows that molecular ions such as TiN and SiN2 are present in
large numbers, which can be related to the covalent nature of the Ti-N and Si-N bonds [13]. In addition, alternating dense and sparse crescents along the length of all reconstructed tips can be seen in cross-sectional images. This phenomenon is commonly occurring in APT analyses and is related to the inability of the reconstruction algorithm to account for bursts of ions originating from a small part of the apex. Such a burst can be produced by correlated evaporation [37], a common phenomenon in ceramic materials [38â40], to which TiSiN belongs [30].
When several ions created in the same pulse are detected it is known as a multiple event. It can be, e.g., from a burst as described in the previous paragraph or through a mid-flight
dissociation of a field evaporated molecular ion. Should they arrive at the detector too close in space and time, not all of them will be detected. Certain elements and phases are more prone to field evaporate in this way than others, which may skew the measured composition. However, elements with a wide isotopic signature, i.e. with several isotopes, are less likely to arrive at the detector simultaneously as the ion speed varies with mass. The multiple events percentage in the present experiment varied between 34 % and 40 %. This is high compared to metals, but that is a common situation with transition metal nitrides [41â43] and it is slightly lower than the values for TiSiN reported by Tang et al. [30]. Because of differences in isotopic signature and field evaporation characteristics, a higher percentage of N is expected to be lost in multiple events compared to Ti and Si), causing an underestimation of N. This is believed to be the largest source of error in the APT-determined composition, with as much as a few percent.
3.4 Compositional analyses
Quantitative compositional analyses are shown here to be made possible in APT of TiSiN by isotopic substitution, as confirmed by using complementary techniques. The compositions as determined by ToF-E ERDA are presented in Table 1. EDS corroborates the Si concentrations and Ti:Si ratios measured by ERDA, as shown in Fig. 3 (a) and (c), while there is a slight discrepancy of 1-3 at. % for Ti and N concentrations, as shown in Fig. 3 (b) and (d). This discrepancy can be attributed to the difficulty in quantifying light elements, such as N, with
EDS. The relative error on the ERDA measurement is ~3 % for Ti and N. For Si it is ~5 % and increases slightly with decreasing Si content.
APT data points in Fig. 3 represent measurements of three different tips of each composition. The error bars account for film inhomogeneity, both within and between the tips, as well as the ranging of each individual mass spectrum. As mentioned in Section 3.2, the peak at 15 Da in Fig. 2 (a) is considered to have two contributions, 15N+ and 15N22+, with a ratio that cannot
be readily determined. The APT data points in Fig. 3 are generated with the 15 Da peak ranged as 15N+ only. This, combined with the multiple events effect described in Section 3.3,
means that N is likely underestimated. Such discrepancies between ERDA and APT can be seen in Fig. 3, but they are generally within the error margin, which in turn suggests that the assumption that most of the 15 Da peak is generated by 15N+ ions is valid.
The 14N content as measured by ERDA, shown as the difference between the orange and
the red curve in Fig. 3 (d) while exact figures are given in Table 1, does not exceed 1 at. % for any film. The purity of the N2 gas is â„99 % and the 15N enrichment â„98 %. Since no sign of 14N-diffusion from the film surface can be found in the ERDA depth profiles (not shown), the
likely source of 14N is the reactive gas. Because 14N is interpreted as Si in APT, an
overestimation of less than 1 at. % Si is to be expected.
H and H2 from the residual chamber gas are here the most prominent contaminations in
APT (<2.5 at. %), but trace amounts (â€0.1 at. %) of C and O are detected in the present analyses by both APT and ERDA. The APT technique probes small volumes, whereas ERDA probes comparatively large areas, but close to the surface. ERDA is thus more strongly
influenced by surface contamination and especially surface roughness, but less influenced by local inhomogeneity (except in the analysis direction, which then is measured). C, O and N can be slightly overestimated in ERDA because of rough sample surfaces.
Fig. 3. Measurements with EDS, ERDA, and APT of (a) x (Si/(Ti+Si)), and the concentration of (b) Ti, (c) Si, and (d) N in at. % as a function of x (Si/(Ti+Si)) on Ti1-xSix15N thin film samples,
as measured by ERDA.
Table 1. The designation and composition of all films as determined by ERDA. The APT analyses show that contaminations in the present samples are homogeneously distributed, except for Ga (<0.5 at. %) from the tip specimen preparation that primarily is found in the ~5 nm closest to the surface of the APT tips. Surface contaminations of the film should not influence APT because the initial part of the measurement was not reconstructed, as this part often contains more implanted Ga+ ions, contaminations from sample transfer,
and the fact that the DC voltage is low, so that heavy ions may require longer flight time than the duration between pulses. The APT compositions of TiSi15N in Fig. 3 have been normalized
to 100 % after excluding all these contaminants.
All in all, the compositional analyses confirm that most Si and N can be distinguished from one another in APT. As discussed, the relatively small discrepancies can be attributed to multiple events, remaining mass spectral overlaps, the ranging of peaks, and film
inhomogeneity (in combination with differences in probed volume inherent to the different techniques). As some discrepancies even out, the difference between the APT and ERDA compositional measurements is found to be even smaller than the differences between APT tips. The compositions measured with ERDA and the mean compositions measured by APT differs with <1.6 at. % for Ti, <0.9 at. % for Si, and <1.5 at. % for 15N, all of which are lower
than the largest deviation from the experimental mean APT value; 1.9 at. %, 1.5 at. %, and 2.2 at. % for Ti, Si, and N, respectively. This shows that APT of TiSiN with isotopic substitution may be used to accurately determine the composition of the material.
3.5 Local composition, voxel size, and delocalization effects
The APT composition considered so far is the relative concentration of different ions in entire analyses. If compositional differences within tips are to be visualized, one cannot directly relate the abundance of ions in two regions unless they have the same volume. To account for this, the ions are sampled into volumes. The volumes can have a fixed number of ions and varying sizes, or a fixed size and varying number of ions. The former will be referred to as âblocksâ, while the latter are called âvoxelsâ for volume pixel.
To reduce sampling artifacts, a 3D Gaussian smoothening known as âdelocalizationâ is applied to each ion. This accounts for uncertainties in the position of the ion and enables ions
Designation Ti0.99Si0.0115N Ti0.95Si0.0515N Ti0.92Si0.0815N Ti0.89Si0.1115N Ti0.85Si0.1515N Ti0.82Si0.1815N Ti0.81Si0.1915N Ti (at. %) 51.4 48.8 47.7 45.5 43.2 41.5 40.7 Si (at. %) 0.5 2.7 3.9 5.7 7.7 9.3 9.8 15N (at. %) 47.2 47.5 47.2 47.8 48.3 48.3 48.6 14N (at. %) 0.8 0.9 1.0 0.8 0.8 0.8 0.8 C (at. %) 0.1 0.1 0.1 0.1 0.0 0.1 0.1 O (at. %) 0.0 0.0 0.1 0.1 0.0 0.0 0.0 x 0.010 0.052 0.076 0.111 0.151 0.183 0.194
to contribute partially to several voxels. The value of the delocalization corresponds roughly to the 3Ï value of the Gaussian smoothening [11,44], and has been chosen to twice the voxel edge length in all directions in this experiment. The default option in IVAS is three times the voxel edge length in the x and y-direction and 1.5 times in the z-direction, but this was deemed too long in the x and y-direction given the small scale of the structures of interest in TiSiN, and an isotropic smoothening was chosen to take care of the influence of correlated evaporation on the certainty of the ion positions in the z-direction.
An overview of the same region from Ti0.81Si0.1915N imaged with different voxel sizes and
delocalization settings is shown in Fig. 4. This gives an indication of the range in which a useful voxel size can be found. Choosing small voxels, e.g., column A in Fig. 4, is beneficial for the spatial resolution, but will lead to high uncertainty in composition due to the low number of ions contributing to each voxel.
Fig. 4. Partial composition maps illustrating how the voxel edge length and delocalization setting of cubic voxels affect the visualization of the nanostructure in Ti0.81Si0.1915N. Each box
is 18 x 29 nm2.
With decreasing number of ions per voxel, i.e. going from column D to column A (and beyond), the information will at some point be indiscernible from noise. On the other hand, if the voxel size is too large, as in column D in Fig. 4, features of interest may instead be interpreted as larger than they are or even become averaged to such an extent that they are indiscernible from their surroundings. Changes in feature size and excessive averaging can also be an effect of a too large delocalization setting, as seen in panel A1. In general terms, this is known as a scale-space problem [45]. The choice of voxel size is not trivial and very important for
representing the data correctly. As such, it should preferably be made in a controlled and repeatable way, thereby ensuring that it correlates to the size of the features of interest. The method for determining a proper block size by Hetherington and Miller [46], previously tested for voxels by Torres et al. [47] was applied to the data in order to determine the most meaningful voxel size. The differences in composition ÎC in this case Si/(Ti+Si)) between neighboring voxels are averaged and the values for different voxel sizes are presented in a log-log plot, where the x-axis shows the mean number of ions per voxel. A plateau in such a plot would suggest that the composition is stable, i.e. that the local composition remains approximately the same even when the voxel size changes within the limits of the plateau. The best compromise between spatial and compositional resolution would then be the smallest voxel size within the plateau region. One such plot for TiSi15N is shown in Fig. 5, which also
contains the same plot for two datasets where the total number of ions, their positions and the average composition are the same, but the elemental identity of the ions has been randomly distributed. As can be seen, the experimental data clearly differs from the randomized data, excluding the possibility of a random solid solution. This was confirmed using frequency distribution analysis (not shown). There is, however, no plateau region with stable local composition for our samples.
Fig. 5. Mean difference in composition ÎC (Si/(Ti+Si)) in Ti0.81Si0.1915N of neighboring voxels
as a function of mean number of ions per voxel. The delocalization 3Ï value was set to twice the voxel edge length.
This result was expected given that the method in question works best for samples with strong compositional differences between the features and their surroundings [46]. The measured local composition will thus always be affected by the choice of voxel size, making it even more important to choose the voxel size in a controlled and repeatable way. Because the XRD diffractograms in Fig. 1 show that the grain size is decreasing to the nanometer scale with increasing Si content, small features are to be expected. A small voxel size is thus needed to resolve these features, but the data must still be discernible from noise.
The mean of the two datasets with randomly redistributed chemical identities simulates the noise level of the particular choice of composition, ion positions, voxel size, and
delocalization setting. The difference between the experimental data and the mean of the randomly re-distributed data as a function of mean number of ions per voxel, which is proportional to the voxel size, is shown in Fig. 6 for all six APT analyses.
Fig. 6. Difference between the experimental mean and randomized mean ÎC (Si/(Ti+Si)) of neighboring voxels (presented for one sample in Fig. 5), as a function of the mean number of ions per voxel. Curves for six APT analyses from two Ti1-xSix15N film compositions, as well as
the mean of all curves, are included. Each dot, square or triangle represents a voxel size starting with 0.53 nm3 and increasing the voxel edge length with 0.1 nm up to 1.2 nm, after
which 0.2 nm is added with each step up to 2.0 nm. The mean number of ions per voxel approximately correlating to cubic voxels with 0.8 nm edges is highlighted.
When the number of ions per voxel becomes too low, i.e. the voxel size becomes too small, the difference approaches zero due to an increasing noise level. When the number of ions per voxel becomes too high, the difference decreases as the features are averaged with their surroundings. The optimum voxel size for visualizing the data is found close to the maximum difference, which would have corresponded with the largest voxel size within the plateau region, should such a region have existed. To increase spatial resolution, a slightly lower voxel size can be tolerated as long as the curve is not too steep and for simplicity in comparing data, a single setting for all tips is desired.
Taking all six atom probe tips into account, a voxel size of 0.83 nm3 is identified as the most
suitable choice in this study. As the value of the delocalization was fixed to twice the voxel edge length, the resulting voxel size and delocalization correspond to panel B3 in Fig. 4. Using these settings of the voxel size and delocalization, the composition can be mapped.
In the optimized composition maps of Fig. 7, the spatial separation of Si-enriched and Ti-enriched domains in two cross-sections perpendicular to the growth and analysis direction can be seen. Fig. 7 (a) shows a Ti0.81Si0.1915N sample while (b) shows a Ti0.92Si0.0815N sample.
similarity of the structure in the two composition maps is interesting, given the difference in grain size, indicated by differences in broadening of the (002) peak in Fig. 1 and would suggest that the grain refinement is primarily an effect of the difference in Si content, rather than its distribution. The small grains believed to cause the broadening of the (002) peak in Fig. 1 are confirmed by the dark field TEM micrograph of a plan-view Ti0.81Si0.1915N sample
with corresponding selected area electron diffraction (SAED) pattern shown in Fig. 7 (c), where a minority of the grains have bright contrast.
Fig. 7. Composition map showing the nano-scale separation of Si and Ti for (a) Ti0.81Si0.1915N,
(b) Ti0.92Si0.0815N as well as (c) dark field TEM micrograph with corresponding SAED pattern
from a plan-view section of the Ti0.81Si0.1915N film. The approximate objective aperture
position and size is marked with a semitransparent circle.
There is no sign of (111) grains at this magnification, but faint signals from (222), (133), and (224) grains as well as strong signals from a few orientations of (002) and (022). When a larger area than that in Fig. 7 (c) contributes to the diffraction pattern, complete rings are formed and a faint signal from (111) is also visible. As the size, rather than the grain
aperture was placed to include parts of several diffraction rings. Both the TEM micrograph and the APT composition maps are plan-views, i.e. recorded with the growth direction as surface normal. The size of single bright domains in Fig. 7 (c) is on average 4.8 nm ± 2.0 nm across, while the domains in the APT composition maps are slightly smaller (4.2 nm ± 2.2 nm for Ti0.81Si0.1915N and 3.9 nm ± 2.6 nm in Ti0.92Si0.0815N), but still in good agreement. We have
thus shown that isotopic substitution with 15N enables local compositional analysis of a TiSiN
material system with APT.
3.6 Criteria for successful application of isotopic substitution
Having demonstrated how to resolve mass-overlaps in the TiSiN system by isotopic
substitution of N, we now turn to consider such prospects for other materials systems. Several criteria must be fulfilled in order to exploit an isotopic substitution.
To begin with, the element whose isotope is to be substituted must have at least two stable isotopes, or two isotopes with sufficiently long half-lives to remain unchanged during the course of the investigation and preferably be safe handle, all of which applies to the TiSiN case. Thereby, some elements such as F, P, As, and Pr are excluded from such consideration since they have only one sufficiently stable isotope. These elements can still be part of the material system, but passively as their peaks cannot be shifted by substitution.
Secondly, separation of the isotopes must be possible and the process must be efficient enough to make the isotope available in quantity and high enough purity at a reasonable price.
Thirdly, the elements of an overlapping peak should preferably have no, or only small, peaks in the vicinity of the overlapping peak. This makes elements with few stable isotopes and/or one predominant isotope more likely to produce useful results.
Lastly, the element should preferably not be a common residual gas in the analysis chamber, i.e. H for the ultra-high vacuum of the atom probe. In the case of H, it is not only because the peaks may overlap with the background peak, but there may also be hydride formation. The former would occur with 1H2 from the residual gas if 2H (deuterium) is used,
but could be remedied by using 3H (tritium), since its half-life is as long as 12.3 years [48].
Some information could also be acquired by, e.g., comparing the peaks at 1 Da to 4 Da in samples grown in pure 1H with those grown in pure 2H [49,50]. Such investigations are
facilitated by voltage pulsing, since less molecular ions are generally formed than with thermal pulsing, thereby minimizing the effects of the H2+ and 2H+ peak overlap. The latter,
hydride formation, would occur after residual 1H adheres to the tip surface and forms
complex hydride ions upon field evaporation. In general, the metal hydride complex requires lower evaporation field than the pure metal, i.e. the hydrides field evaporate easier [51]. However, these molecular ions are primarily detected when the field is low or the element has a high affinity for H (e.g., Zr or Si) [1,14], as they are prone to further field ionization and subsequent field dissociation [51].
Another matter of importance for successful isotopic substitution is the charge state of the substituted ion, as the length of the peak shift is the atomic mass difference of the isotopes divided by the charge state. Since the shift will be at most a couple of Da for stable isotopes,
the usefulness of the method will vary from system to system. Additionally, if the peaks involved are broadened by low mass resolution or have significant tails, the result will be influenced by the partial overlap of the remaining and shifted peaks.
In summary thus far, factors such as availability and stability of suitable isotopes, severity of the overlap, charge state of the substituted ion, and mass resolution will determine whether an isotopic substitution is viable and indeed beneficial. In the case of TiSiN, N has two isotopes that are possible to separate while Si has three. All five are commercially available, but
isotope-enriched Ti:Si compound cathodes with different Ti:Si ratios are less common, which is why 15N-enriched gas was chosen for this study. By moving the overlap from 14 Da to
15 Da, the overlapping Si peak will be at 3.1 at. % of the natural abundance, rather than
92.2 at. %, which is a significant improvement. The mass resolution of the instrument with the experimental parameters used in this study was found to be high enough to resolve all peaks from predominant isotopes after the 0.5-1 Da shift of the N peaks.
3.7 Materials systems with isotopic-resolvable mass overlaps in APT
Next, we consider cases for other technologically important materials system that suffer from mass overlap and where APT analysis would benefit from isotopic substitutions. Obviously from the present results, systems containing both Si and N require isotopic substitution of 14N
if time-of-flight techniques are of interest. Nitrogen-doped Si and SiC are two such candidates used in electronic applications, where the demand for exact figures of, e.g., purity and dopant concentrations is necessary to ensure good device performance. Many complex devices are small enough to fit into an APT tip [52], making it a powerful instrument for linking the performance of the device to particular structures.
Other semiconductors, such as Ge-doped GaAs or InP, are also problematic to analyze using time-of-flight techniques because of mass spectral overlaps. Owing to its many stable
isotopes, natGe can be substituted to reduce the overlap to ~0 % for both GaAs and InP. The
best choice of isotope for APT analysis found in literature would be 70Ge-enriched to
99.99 % [53] or 74Ge-enriched to ~99.93 at. % [54]. Both would decrease the risk of partial
overlap with Ga (from the GaAs, specimen preparation or both). Neither of these two would overlap with 69GaAs, 71GaAs, 113InP, or 115InP.
Another element that is suitable for substitution or as a passive component is B. Its natural abundance is somewhat similar to that of N in that they both have two stable isotopes of which one is predominant. However, the ratio of the two peaks differs between the elements; the percentages are 19.9 at. % and 80.1 at. % for 10B and 11B respectively, compared to
99.6 at. % and 0.4 at. % for 14N and 15N, respectively. Even if full peak separation cannot be
achieved, isotopic substitution would still improve the reliability of peak deconvolution
calculations and allow the ion-specific loss of ions due to multiple events to be investigated. In addition, 10B-enriched B4C is already used commercially in neutron absorbing applications,
such as nuclear control rods [55] and neutron detectors [56], and is thereby available in bulk. As regulators of the fission rate in nuclear power plants, the control rods must remain intact and functional at all times. B4C pellet-containing nuclear control rods may exhibit faster
performance of the rods remains unchanged and the use of cladding should prevent damages to the structure of the control rod itself, the structural integrity of the B4C pellets themselves
could be important for safe use in future applications. An isotopic substitution to pure 10B4C
would not yield perfect differentiation from Na because of the 10B13C-ion, but should the
chemical sensitivity be deemed too low, a double isotopic substitution to 10B12C is possible,
since isotope-enriched C is commercially available. Double isotopic substitutions have previously been used in e.g., infrared absorption spectrometry [58,59] and FT-Raman spectroscopy [60], but have not been reported for APT to date.
Cr is added to steels in order to increase the oxidation and corrosion resistance, i.e. to make it stainless. A complete overlap will occur between the heaviest of the four stable Cr isotopes,
54Cr2+ (2.366 at. %), and the lightest of the four stable Fe isotopes, 54Fe2+ (5.845 at. %) at 27
Da. Since the mass difference is as small as 0.00037 Da and the charge state is the same, even kinetic energy analysis of the ions would not be able to separate them [14]. An additional element with mass spectral peak at 27 Da is Al, which is a common alloying agent in FeCr. Single or double isotopic substitution, for FeCr or FeCrAl respectively, is possible and goes to show that kinetic energy analysis and isotopic substitution can be regarded as
complementary techniques. However, the FeCrAl overlap has recently been spatially resolved using single-ion deconvolution by London et al. [61] without the need for isotopic
substitution. This method could be used on TiSinatN as well, although London et al. state that
the method is best applied when there is a spatial variation of the mass spectral overlap [61], which is not the case for the films studied in this paper.
Isotopic substitution applied to these materials systems, and many more not covered by this paper, would increase the information obtained from APT analysis and help solve outstanding questions. It could also be used instead of, or together with, kinetic energy analysis.
4. Conclusions
Isotopic substitution with 15N has been successfully conducted in TiSiN thin films, effectively
enabling the differentiation of Si and N in APT. The number of ions attributed to erroneous elements has been significantly reduced to yield a Si concentration accuracy within 1.2 at. % from the value measured by ERDA, while the Ti and N concentration was within 1.6 at. % and 2.1 at. % from the ERDA values, respectively. Factors such as ranging, multiple events, and remaining overlaps ultimately limit the accuracy of the compositional determination.
The algorithm for finding the appropriate block size by Hetherington and Miller [46] was adapted to voxels and applied to the experimental APT data, but did not yield a range where the local composition is not affected by the choice of voxel size. However, by comparing the experimental data with itself after randomly re-distributing the chemical identities of the ions, a voxel size of 0.83 nm3 was found to be the best choice for balancing noise and excessive
averaging. This notably small voxel size was then used for visualizing the experimental data, in which Si segregation on the nanometer scale was evident. This finding suggests further investigation of the material is warranted.
Criteria for identifying materials systems with properties promising for isotopic substitution have been established and applied to the TiSiN system. These are isotope stability and availability, simple isotopic signature, and elements not prominent in the residual gas. We also identified a number of other systems of scientific and commercial interest, and proposed them for future APT analysis with isotopic substitution. Isotopic substitution is thus a viable option for analyzing TiSiN as well as other materials systems.
Acknowledgements
The financial support of the VINN Excellence Center on Functional Nanoscale Materials (FunMat) 2007-00863, the Swedish Research Council (VR) project grant 2013-4018, the Swedish Government Strategic Research Area Grant in Materials Science (Grant SFO Mat-LiU 2009-00971) on Advanced Functional Materials, and Knut and Alice Wallenberg Project
Isotope is greatly appreciated. We are also thankful for the access to the Tandem Laboratory,
Uppsala University, for the ERDA measurements.
References
[1] B. Gault, M.P. Moody, J.M. Cairney, S.P. Ringer, Atom Probe Microscopy, Springer, New York, 2012. doi:10.1007/978-1-4614-3436-8.
[2] R.G. Forbes, Field electron and ion emission from charged surfaces: a strategic historical review of theoretical concepts, Ultramicroscopy. 95 (2003) 1â18. doi:10.1016/S0304-3991(02)00293-0.
[3] D.R. Kingham, The post-ionization of field evaporated ions; a theoretical explanation of multiple charge states, Surf. Sci. 116 (1982) 273â301.
doi:10.1016/0167-2584(82)90179-7.
[4] P. Bas, A. Bostel, B. Deconihout, D. Blavette, A general protocol for the reconstruction of 3D atom probe data, Appl. Surf. Sci. 87â88 (1995) 298â304.
doi:10.1016/0169-4332(94)00561-3.
[5] N. Mayama, C. Yamashita, M. Owari, T. Kaito, M. Nojima, Stress of needle specimen on the three-dimensional atom probe (3DAP), Surf. Interface Anal. 40 (2008) 1610â1613. doi:10.1002/sia.2905.
[6] C. Moy, G. Ranzi, T.C. Petersen, S. Ringer, Macroscopic electrical field distribution and field-induced surface stresses of needle-shaped field emitters, Ultramicroscopy. 111 (2011) 397â404. doi:10.1016/j.ultramic.2011.01.024.
[7] J.H. Bunton, J.D. Olson, D.R. Lenz, T.F. Kelly, Advances in pulsed-laser atom probe: Instrument and specimen design for optimum performance, Microsc. Microanal. 13 (2007) 418â427. doi:10.1017/S1431927607070869.
[8] A. Cerezo, A. Gomberg, G.D.W. Smith, P.H. Clifton, Aspects of the performance of a
femtosecond laser-pulsed 3-dimensional atom probe, Ultramicroscopy. 107 (2007) 720â 725. doi:10.1016/j.ultramic.2007.02.025.
[9] G.L. Kellogg, T.T. Tsong, Pulsed-laser atom-probe field-ion microscopy, J. Appl. Phys. 51 (1980) 1184â1193. doi:10.1063/1.327686.
[10] L.J. Lauhon, P. Adusumilli, P. Ronsheim, P.L. Flaitz, D. Lawrence, Atom-Probe Tomography of Semiconductor Materials and Device Structures, MRS Bull. 34 (2009) 738â743.
doi:10.1557/mrs2009.248.
[11] M.K. Miller, R.G. Forbes, Atom-Probe Tomography: The Local Electrode Atom Probe, Springer US, Boston, MA, 2014. doi:10.1007/978-1-4899-7430-3.
[12] T.T. Tsong, Field penetration and band bending for semiconductor of simple geometries in high electric fields, Surf. Sci. 85 (1979) 1â18.
[13] T.J. Prosa, R.A. Alvis, T.F. Kelly, Observations of cluster ions originating from non-traditional atom probe materials, Microsc. Microanal. 14 (2008) 1236â1237. doi:10.1017/S1431927608083736.
[14] T.F. Kelly, Kinetic-Energy Discrimination for Atom Probe Tomography, Microsc. Microanal. 17 (2011) 1â14. doi:10.1017/S1431927610094468.
[15] L. Shizhi, S. Yulong, P. Hongrui, Ti-Si-N films prepared by plasma-enhanced chemical vapor deposition, Plasma Chem. Plasma Process. 12 (1992) 287â297.
doi:10.1007/BF01447027.
[16] S. VepĆek, S. Reiprich, L. Shizhi, Superhard nanocrystalline composite materials: the TiN/Si3N4 system, Appl. Phys. Lett. 66 (1995) 2640â2642. doi:10.1063/1.113110.
[17] F. Vaz, L. Rebouta, S. Ramos, M.F. da Silva, J.C. Soares, Physical, structural and mechanical characterization of Ti1-xSixNy films, Surf. Coat. Technol. 108â109 (1998) 236â240.
doi:10.1016/S0257-8972(98)00620-3.
[18] M. Diserens, J. Patscheider, F. LĂ©vy, Mechanical properties and oxidation resistance of nanocomposite TiNâSiNx physical-vapor-deposited thin films, Surf. Coat. Technol. 120â
121 (1999) 158â165. doi:10.1016/S0257-8972(99)00481-8.
[19] A. Flink, T. Larsson, J. Sjölén, L. Karlsson, L. Hultman, Influence of Si on the
microstructure of arc evaporated (Ti,Si)N thin films; evidence for cubic solid solutions and their thermal stability, Surf. Coat. Technol. 200 (2005) 1535â1542.
doi:10.1016/j.surfcoat.2005.08.096.
[20] E.V. Shalaeva, S.V. Borisov, O.F. Denisov, M.V. Kuznetsov, Metastable phase diagram of Ti-Si-N(O) films (CSi < 30 at.%), Thin Solid Films. 339 (1999) 129â136.
doi:10.1016/S0040-6090(98)01259-0.
[21] S. VepĆek, M.G.J. VepĆek-Heijman, P. Karvankova, J. Prochazka, Different approaches to superhard coatings and nanocomposites, Thin Solid Films. 476 (2005) 1â29.
doi:10.1016/j.tsf.2004.10.053.
[22] L. Hultman, A. Flink, J. Bareño, V. Petrova, J.E. Greene, I. Petrov, H. Söderberg, M. Odén, K. Larsson, Interface structure in superhard TiN-SiN nanolaminates and nanocomposites: Film growth experiments and ab initio calculations, Phys. Rev. B. 75 (2007) 155437. doi:10.1103/PhysRevB.75.155437.
[23] R.F. Zhang, S. VepĆek, A.S. Argon, Understanding why the thinnest SiNx interface in
transition-metal nitrides is stronger than the ideal bulk crystal, Phys. Rev. B. 81 (2010) 155437. doi:10.1103/PhysRevB.81.245418.
[24] F. Tang, B. Gault, S.P. Ringer, P. Martin, A. Bendavid, J.M. Cairney, Microstructural investigation of TiâSiâN hard coatings, Scr. Mater. 63 (2010) 192â195.
doi:10.1016/j.scriptamat.2010.03.050.
[25] M. Thuvander, G. Ăstberg, L.K.L. Falk, M. Ahlgren, Atom probe tomography of a Ti-Si-Al-C-N coating grown on a cemented carbide substrate, Ultramicroscopy. 159 (2014) 308â 313. doi:10.1016/j.ultramic.2015.04.008.
[26] W. Sha, L. Chang, G.D.W. Smith, L. Cheng, E.J. Mittemeijer, Some aspects of atom-probe analysis of FeC and FeN systems, Surf. Sci. 266 (1992) 416â423. doi:10.1016/0039-6028(92)91055-G.
[27] T. Kinno, M. Tomita, T. Ohkubo, S. Takeno, K. Hono, Laser-assisted atom probe
tomography of 18O-enriched oxide thin film for quantitative analysis of oxygen, Appl.
Surf. Sci. 290 (2014) 194â198. doi:10.1016/j.apsusc.2013.11.039.
[28] T. Kinno, K. Kitamoto, S. Takeno, M. Tomita, Laser-assisted atom probe tomography of
15N-enriched nitride thin films for analysis of nitrogen distribution in silicon-based
structure, Appl. Surf. Sci. 349 (2015) 89â92. doi:10.1016/j.apsusc.2015.04.200. [29] K. Thompson, D. Lawrence, D.J. Larson, J.D. Olson, T.F. Kelly, B. Gorman, In situ
site-specific specimen preparation for atom probe tomography, Ultramicroscopy. 107 (2007) 131â139. doi:10.1016/j.ultramic.2006.06.008.
[30] F. Tang, B. Gault, S.P. Ringer, J.M. Cairney, Optimization of pulsed laser atom probe (PLAP) for the analysis of nanocomposite TiâSiâN films, Ultramicroscopy. 110 (2010) 836â843. doi:10.1016/j.ultramic.2010.03.003.
[31] M. Freimer, G. Kollia, G.S. Mudholkar, C.T. Lin, A Study of the Generalized Tukey Lambda Family, Commun. Stat. A-Theor. 17 (1988) 3547. doi:10.1080/03610928808829820. [32] H.J. Whitlow, G. Possnert, C.S. Petersson, Quantitative Mass and Energy Dispersive Elastic
Recoil Spectrometry: Resolution and Efficiency Considerations, Nucl. Instrum. Meth. B. 27 (1987) 448â457. doi:10.1016/0168-583X(87)90527-1.
[33] M.S. Janson, CONTES Conversion of Time-Energy Spectra - a Program for ERDA Data Analysis, Internal Report, Uppsala University, 2004.
[34] G. Greczynski, J. Patscheider, J. Lu, B. Alling, A. Ektarawong, J. Jensen, I. Petrov, J.E. Greene, L. Hultman, Control of Ti1-xSixN nanostructure via tunable metal-ion momentum transfer
during HIPIMS/DCMS co-deposition, Surf. Coat. Technol. 280 (2015) 174â184. doi:10.1016/j.surfcoat.2015.09.001.
[35] Z. Wu, C. Wu, X. Liu, Y. Deng, Q. Gong, D. Song, H. Su, Double Ionization of Nitrogen from Multiple Orbitals, J. Phys. Chem. A. 114 (2010) 6751â6756. doi:10.1021/jp1018197. [36] J.K. Bohlke, J.R. de Laeter, P. De Bievre, H. Hidaka, H.S. Peiser, K.J.R. Rosman, P.D.P. Taylor,
Isotopic compositions of the elements, 2001, J. Phys. Chem. Ref. Data. 34 (2005) 57â67. doi:10.1063/1.1836764.
[37] F. De Geuser, B. Gault, A. Bostel, F. Vurpillot, Correlated field evaporation as seen by atom probe tomography, Surf. Sci. 601 (2007) 536â543. doi:10.1016/j.susc.2006.10.019. [38] M.K. Miller, Three-dimensional atom probes, J. Microsc.-Oxford. 186 (1997) 1â16.
doi:10.1046/j.1365-2818.1997.1910759.x.
[39] M. Thuvander, A. Kvist, L.J.S. Johnson, J. Weidow, H.-O. Andren, Reduction of multiple hits in atom probe tomography, Ultramicroscopy. 132 (2013) 81â85.
doi:10.1016/j.ultramic.2012.12.005.
[40] D.W. Saxey, Correlated ion analysis and the interpretation of atom probe mass spectra, Ultramicroscopy. 111 (2011) 473â479. doi:10.1016/j.ultramic.2010.11.021.
[41] L.J.S. Johnson, M. Odén, L. Hultman, M. Thuvander, K. Stiller, Spinodal decomposition of Ti0.33Al0.67N thin films studied by atom probe tomography, Thin Solid Films. 520 (2012)
4362â4368. doi:10.1016/j.tsf.2012.02.085.
[42] I. Povstugar, P.-P. Choi, D. Tytko, J.-P. Ahn, D. Raabe, Interface-directed spinodal decomposition in TiAlN/CrN multilayer hard coatings studied by atom probe
tomography, Acta Mater. 61 (2013) 7534â7542. doi:10.1016/j.actamat.2013.08.028. [43] J. Angseryd, F. Liu, H.-O. Andren, M. Thuvander, S.S.A. Gerstl, Quantitative APT analysis of
Ti(C,N), Ultramicroscopy. 111 (2011) 609â614. doi:10.1016/j.ultramic.2011.01.031. [44] D.J. Larson, T.J. Prosa, R.M. Ulfig, B.P. Geiser, T.F. Kelly, Local Electrode Atom Probe
Microscopy: A Userâs Guide, Springer, New York, 2013. doi:10.1007/978-1-4614-8721-0. [45] T. Lindeberg, Scale Selection Properties of Generalized Scale-Space Interest Point
[46] M.G. Hetherington, M.K. Miller, Some Aspects of the Measurement of Composition in the Atom Probe, J. Phys. Colloq. 50 (1989) C8-535. doi:10.1051/jphyscol:1989892.
[47] K.L. Torres, M. Daniil, M.A. Willard, G.B. Thompson, The influence of voxel size on atom probe tomography data, Ultramicroscopy. 111 (2011) 464â468.
doi:10.1016/j.ultramic.2011.01.001.
[48] L.L. Lucas, M.P. Unterweger, Comprehensive review and critical evaluation of the half-life of tritium, J. Res. Natl. Inst. Stand. Technol. 105 (2000) 541â549.
doi:10.6028/jres.105.043.
[49] D. Haley, S.V. Merzlikin, P. Choi, D. Raabe, Atom probe tomography observation of hydrogen in high-Mn steel and silver charged via an electrolytic route, Int. J. Hydrogen Energ. 39 (2014) 12221â12229. doi:10.1016/j.ijhydene.2014.05.169.
[50] G. Sundell, M. Thuvander, A.K. Yatim, H. Nordin, H.-O. AndrĂ©n, Letter: Direct observation of hydrogen and deuterium in oxide grain boundaries in corroded Zirconium alloys, Corros. Sci. 90 (2015) 1â4. doi:10.1016/j.corsci.2014.10.016.
[51] Z.M. Stepien, T.T. Tsong, Formation of metal hydride ions in low-temperature field evaporation, Surf. Sci. 409 (1998) 57â68. doi:10.1016/S0039-6028(98)00200-3.
[52] T.F. Kelly, D.J. Larson, K. Thompson, R.L. Alvis, J.H. Bunton, J.D. Olson, B.R. Gorman, Atom probe tomography of electronic materials, Annu. Rev. Mater. Res. 37 (2007) 681â727. doi:10.1146/annurev.matsci.37.052506.084239.
[53] V.I. Ozhogin, A.V. Inyushkin, A.N. Taldenkov, A.V. Tikhomirov, G.Ă. Popov, E. Haller, K. Itoh, Isotope effect in the thermal conductivity of germanium single crystals, JETP Lett+. 63 (1996) 490â494. doi:10.1134/1.567053.
[54] M.F. Churbanov, A.V. Gusev, A.D. Bulanov, A.M. Potapov, Monoisotopic varieties of silicon and germanium with a high chemical and isotopic purity, Russ. Chem. B+. 62 (2013) 270â275. doi:10.1007/s11172-013-0040-2.
[55] P. Manoravi, M. Joseph, N. Sivakumar, P.R.V. Rao, Quasi-non-destructive isotopic ratio measurement of boron in irradiated control rod B4C pellets using a home-built reflectron
time-of-flight mass spectrometer, Int. J. Mass Spectrom. 309 (2012) 148â153. doi:10.1016/j.ijms.2011.09.012.
[56] C. Höglund, J. Birch, K. Andersen, T. Bigault, J.-C. Buffet, J. Correa, P. van Esch, B. Guerard, R. Hall-Wilton, J. Jensen, A. Khaplanov, F. Piscitelli, C. Vettier, W. Vollenberg, L. Hultman, B4C thin films for neutron detection, J. Appl. Phys. 111 (2012) 104908.
doi:10.1063/1.4718573.
[57] A.L. Pitner, Liquid Metal Reactor Absorber Technology, Westinghouse Hanford Company, Richland, Washington, 1990. WHC-SA-0976-FP.
[58] M. Bouchacourt, F. Thevenot, The properties and structure of the boron carbide phase, J. Less-Common Met. 82 (1981) 227â235. doi:10.1016/0022-5088(81)90223-X.
[59] H. Stein, T. Aselage, D. Emin, Infrared absorption in boron carbides: dependence on isotopes and carbon concentration, AIP Conf. Proc. 231 (1991) 322â325.
doi:10.1063/1.40844.
[60] H. Werheit, H.W. Rotter, F.D. Meyer, H. Hillebrecht, S.O. Shalamberidze, T.G. Abzianidze, G.G. Esadze, FT-Raman spectra of isotope-enriched boron carbide, J. Solid State Chem. 177 (2004) 569â574. doi:10.1016/j.jssc.2003.04.004.
[61] A.J. London, D. Haley, M.P. Moody, Single-Ion Deconvolution of Mass Peak Overlaps for Atom Probe Microscopy, Microsc. Microanal. 23 (2017) 300â306.
