Annual Review of Analytical Chemistry
Gas Cluster Ion Beams for Secondary Ion Mass Spectrometry
Nicholas Winograd
Department of Chemistry, Pennsylvania State University, University Park, Pennsylvania 16802, USA; email: [email protected]
Annu. Rev. Anal. Chem. 2018. 11:29–48
Keywords
First published as a Review in Advance on February 28, 2018
bioimaging, cluster ion beams, phospholipids, instrumentation, molecular depth profiling, molecular dynamics computer simulations
The Annual Review of Analytical Chemistry is online
at anchem.annualreviews.org
Abstract
https://doi.org/10.1146/annurev-anchem- 061516-045249
Gas cluster ion beams (GCIBs) provide new opportunities for bioimaging and molecular depth profiling with secondary ion mass spectrometry (SIMS). These beams, consisting of clusters containing thousands of particles, initiate desorption of target molecules with high yield and minimal fragmentation. This review emphasizes the unique opportunities for implementing these sources, especially for bioimaging applications. Theoretical aspects of the cluster ion/solid interaction are developed to maximize conditions for successful mass spectrometry. In addition, the history of how GCIBs have become practical laboratory tools is reviewed. Special emphasis is placed on the versatility of these sources, as size, kinetic energy, and chemical composition can be varied easily to maximize lateral resolution, hopefully to less than 1 micron, and to maximize ionization efficiency. Recent examples of bioimaging applications are also presented.
- c
- Copyright ꢀ 2018 by Annual Reviews.
All rights reserved
ANNUAL REVIEWS
Further
Click here to view this article's online features:
• Download figures as PPT slides • Navigate linked references • Download citations • Explore related articles • Search keywords
29
INTRODUCTION
Since the discovery by J.J. Thompson over 100 years ago that charged particles could be emitted from surfaces (1), secondary ion mass spectrometry (SIMS) has become a mainstay for materials characterization. Currently, the technique is unique in its ability to provide surface-specific molecular information, to acquire spatially resolved mass spectra with submicron resolution, and to acquire in-depth composition of molecular solids with nanometer resolution. The imaging modality is particularly interesting. Originally developed in the SIMS community using focused atomic ion beams to define position (2), the method provides unprecedented chemically specific information at each pixel. Moreover, by stacking images acquired at different sample depths, three-dimensional (3D) information is forthcoming. The SIMS community has retained an unusual buoyancy through the decades, and the recent emergence of cluster ion beams to initiate desorption has had a remarkable effect. With multiple atoms comprising the primary ion, chemical damage to the sample is reduced because each atom carries a smaller share of the incident kinetic energy. Molecular fragmentation is reduced during the desorption process, resulting in nearly fragment-free mass spectra. Molecular depth profiling of complex materials has allowed subsurface information to be acquired for the first time. When combined with imaging, 3D information on the nanoscale is possible to achieve. Cluster projectiles have placed molecular SIMS spectral capabilities, when the m/z < 3,000, nearly on par with other stimulated desorption techniques that do not have these unique aspects (3–5). An excellent comprehensive treatise on this topic has recently been published (6).
SIMS: secondary ion
mass spectrometry, in which secondary ions are generated from primary ion bombardment
Although countless cluster projectile varieties have been proposed, evaluated, and espoused, the implementation of gas cluster ion beams (GCIBs) has a special allure. These clusters, created during supersonic expansion, generate clusters that consist of ∼1,000 to >10,000 component atoms or molecules. To allow for beam focusing and to provide sufficient kinetic energy to initiate molecular desorption, these beams are typically accelerated to >10 keV of kinetic energy. Because kinetic energy divides equally among constituents (7), the kinetic energy of each particle is reduced to a value comparable to chemical bond strengths. Hence, molecular fragmentation, subsurface damage, and interlayer mixing are reduced. There is a fundamental difference between GCIBs
+
and smaller cluster projectiles such as Bi3 or C60+, where the kinetic energy per atom ranges from a few hundred to a few thousand electron volts.
There are practical challenges associated with GCIBs that prevent them from dominating
SIMS laboratories. These beams are characterized by a distribution of cluster sizes that renders pulsing for time-of-flight (TOF)-SIMS problematic. Focusing to a submicron spot has not yet been routinely demonstrated, compromising the ability to acquire high spatial resolution chemical images. Moreover, ionization efficiency is often poor even though molecular desorption is highly efficient. Because of these issues, most SIMS researchers utilize the GCIB as an effective erosion source for molecular depth profiling but employ the smaller cluster sizes for spectral characterization and imaging. As it turns out, however, none of the aforementioned challenges is without solution. Instrumental developments now allow high-quality mass spectra to be acquired easily with GCIBs, and sources focused to at least one micron are becoming available. The great flexibility of tuning the chemistry of the GCIB also has implications for enhancing ionization. In this review, prospects for the exclusive use of GCIBs as the ion source of choice for molecular SIMS experiments are considered by examining each of these factors in detail.
THEORETICAL CONSIDERATIONS
The emergence of GCIB technology has enormously expanded the space available for projectile choice. Variables include size, composition, energy, and incidence angle. How can we optimize
- 30
- Winograd
the experiment? Several important properties need to be considered. First, the sample needs to receive sufficient kinetic energy to initiate molecular desorption. Second, the kinetic energy must be low enough so as not to initiate collision-induced fragmentation. Third, there must be sufficient energy provided to the system to allow ionization during desorption when precharged species are not present. If there are no ions, there is no signal. Finally, because 3D imaging is an important goal, the depth resolution during erosion needs to be optimized by tuning either the kinetic energy or the angle of incidence. The basic challenge is, then, how to find the sweet spot that maximizes all of these demands. Is there enough parameter space to find some set of optimized conditions? Can there be a theory that leads us in the right direction?
To begin, a basic picture of the cluster-solid interaction is needed. This type of picture is most vividly provided by molecular dynamics computer simulations, which for several decades have provided molecular-level information about the energy dissipation process (8–12). An example that illustrates the unique aspect of the GCIB when compared to other atomic or cluster ion projectiles is shown in Figure 1. In this case, the substrate is an Ag(111) crystal, and the projectile is accelerated to a kinetic energy of 15 keV (13) (Ar872 at 15 keV moves at 104 m/s). Hence, the energy per atom varies from 15,000 eV for Ga to 17.2 eV for Arn, where n is the number of atoms in the cluster; in this case, it is 872. Note that both Ga and Au3 penetrate deeply into the Ag crystal, leaving considerable damage and atomic mixing. The displacements are characterized by binary collisions producing a collision cascade of atoms. For C60, the impact results in the formation of a well-defined crater, much like that of a meteor striking the earth. This mesoscale phenomenon is not well described by a collision cascade because the diameter of the C60 molecule (0.7 nm) is larger than the interatomic spacing of the Ag crystal (0.23 nm). This observation has led to the development of analytical theories of particle emission based upon mesoscale fluid flow that are successful for predicting a range of parameters (14, 15). For the GCIB bombardment, the crater is of similar diameter to C60, although the depth is much lower. In addition, Ar872 is large enough to block emission of substrate atoms. Hence, most of the particles are emitted from the crater edge at off-normal angles. The yield of substrate atoms is approximately one-third that of C60, presumably due to this blocking effect and to the lower kinetic energy per particle in the cluster, E/n. It is this parameter that may also affect ionization probability, which is generally lower for GCIBs than for C60. Many other graphical pictures and videos, acquired using molecular dynamics, have been published over the last several years (9–12, 16–20).
With all of the variables associated with GCIBs, and with the possibility of generating clusters with different chemistry (e.g., C60), there has been an effort to find universal relationships that allow for a predictive model using a single equation. A successful approach involves plotting experimentally measured Y/n versus E/n for Ar–GCIBs. Here, Y is the yield of neutral material from a molecular solid in volume units (nm3), and E is the kinetic energy of the cluster (21–25). Such a plot is shown for a variety of materials in Figure 2. Note that the value of Y/n can vary over four orders of magnitude for organic solids and polymers, and even more for atomic solids such as Si, SiO2, and Au. The plot shows that different classes of materials tend to cluster around specific regions. In general, the yield from atomic solids can be an order of magnitude lower than the yield from molecular solids, primarily due to cohesive energy differences. This effect makes the characterization of hybrid materials consisting of metal and organic components challenging due to differential sputtering effects. After classifying the experimental data, the information can be fitted with an empirical equation, which then has some predictive value. From this log–log plot, the Y/n value has a higher slope below 10 eV/atom than above this value. This effect is due to the fact that at higher E/n, the energy cannot efficiently flow away from the cluster impact site fast enough, so the yield of material begins to level off with increasing E. Finally, it has recently been shown that by plotting Y(E/U0), where U0 is the binding energy of the solid, versus E/n, the
•
0 ps
Au3
- C60
- Ar872
Ga
3 ps 29 ps
- 21
- 78
- 317
- 106
1 nm
Figure 1
Molecular dynamics computer simulations for primary ions of varying sizes bombarding an Ag(111) crystal at normal incidence and at a kinetic energy of 15 keV. The layers are color coded to aid visual clarity (13). The first three rows show a cross section of the crystal, whereas the bottom row shows a top-down view to gain insight into the size of the damaged area. The gray particles begin at a depth of 4.6 nm. The numbers in the bottom left corner of the third row are approximate values for the number of Ag atoms ejected for each primary particle impact. The first row shows the initial state of the crystal, and the second and third rows illustrate the time evolution of the damage after 3 ps and 29 ps. Figure adapted with permission from Reference 13. Copyright 2013, SurfaceSpectra Ltd.
behavior of different materials classes can be made to converge (25). The goal of attaining some sort of universally applicable result has nearly been achieved.
The previous discussion applies when the mass of the atom in the cluster does not change. The situation is less clear when trying to compare the behavior of C60 to Ar1000, for example (26, 27). These comparisons can be made by plotting the yield per nucleon of sample versus the energy per nucleon of the projectile (27). This result illustrates that the mass of the projectile is a key parameter. It is interesting that as the energy per nucleon is increased from 0.5 to 200, the ejected mass per nucleon increases by a factor of 200.
Collision-induced fragmentation of target molecules must also be considered when seeking to optimize the size and energy of the GCIB. Using molecular dynamics computer simulations for Ar clusters, Garrison and coworkers (28) have shown that fragmentation begins at ∼50 eV/atom for
- 32
- Winograd
100
10
Irganox 1010
SiO2
HTM-1
1.0
PMMA
Polycarbonate
Polystyrene
0.1
Au
0.01
0.001
0.0001
0.00001
Si
- 1
- 10
- 100
E/n, eV/atom
Figure 2
A log–log plot of Y/n versus E/n (n is the number of atoms in the cluster) for a wide variety of materials bombarded by Ar clusters of various sizes at a 45◦ angle of incidence. Solid lines are the predictions for the universal curve, proposed by Seah (24), whereas the other lines show the behavior of other materials as indicated. Note that the sputtering rate for organic material is 2–4 orders of magnitude greater than for elements or inorganic compounds. Moreover, the sputtering rate for a specific species can vary by more than two orders of magnitude, depending upon the E/n value. Figure adapted with permission from Reference 24.
small molecules such as benzene and octane, and it begins at ∼10 eV/atom with longer entwined molecules such as β-carotene.
In the laboratory, GCIB clusters varying in size from several atoms to 10,000 atoms or more have been synthesized. The yield of emitted secondary ions is found to be a function of both the cluster size and the kinetic energy. How can we best organize all these variables to find the best conditions for SIMS? The heart of the problem is illustrated in Figure 3, where the SIMS spectrum is examined as a function of cluster size. The results show that as the size is increased, the magnitude of the signal decreases in an exponential fashion. However, the number of fragments relative to the number of molecular ions also decreases. A clear example of this effect has been reported using a thin amino acid film as a model compound (29). An interesting dichotomy is then apparent from Figure 3, where the ionization yield drops exponentially with decreasing E/n beginning at ∼20 eV/atom. Maximization of sensitivity also introduces the risk of opening the molecule fragmentation channel. However, in the figure, there appears to be a sweet spot in the 5–10 eV/atom range, where fragmentation is not too severe, and the ion yield drop-off is not too significant.
MOLECULAR DEPTH PROFILING
As noted in the Introduction, an important and unique aspect of cluster SIMS is the ability to interrogate molecular thin films in depth by eroding the sample with the primary ion beam (30). This property is remarkable because this beam creates damage to the surface and subsurface of the molecular film via collision-induced processes, altering the chemical composition. In fact, during the early years of SIMS research using atomic ion beams, so much damage was created that the ion
Static SIMS: method
in which the dose is
dose had to be kept below 1% of the number of surface molecules to acquire enough data to even kept below damage
threshold
produce a mass spectrum, a modality referred to as static SIMS (31). For many years, however, it has been recognized that the degree of damage is much less when employing cluster ion beams,
•
104
- a
- b
103 102
10
1.0 0.1
1
- 500
- 1,000
- 1,500
- 2,000
- 5
- 10
- 15
- 20
- Cluster size, n
- E/n, eV
Figure 3
+
Response of a generic organic molecule to 5-keV Arn bombardment. These graphs illustrate the conundrum associated with molecular ion yield and molecular fragmentation. Panel a shows the typical variation of the molecular ion yield as the cluster size is increased from Ar200 to Ar2000. Panel b shows how the fragments for a variety of molecules change with cluster size. Note that both ordinates are plotted on a log scale and that values of E/n (n is the number of atoms in the cluster) between 5 eV and 10 eV tend to maximize sensitivity and minimize fragmentation. The kinetic energy of the cluster is held constant at 5 keV in all cases. Figure adapted from Reference 29.
similar to what has been observed when diluting the molecule into a matrix. There are at least two factors at play. There is a nonlinear increase in the sputtering yield when the sample is hit by many particles simultaneously, and the lower E/n value associated with clusters inherently yields less damage. The result is attainment of a steady-state composition where the damage created by the primary cluster ion beam is efficiently removed and does not accumulate as the dose is increased.
Many different types of cluster beams, consisting of various sizes and composition, have been espoused over the last several decades. Early experiments using giant glycerol clusters (32), carbon clusters (33), and SF5+ (7) were particularly influential in demonstrating the advantages of cluster SIMS over the use of atomic projectiles. For the first time, it was shown that molecular ion signals persisted as the sample was eroded, and the mass spectra exhibited much less fragmentation than with atomic ion beams. Acceptance was slow, however, due to technical challenges and the fact that these beams could not be focused very well. Without the focusing, the imaging modality so crucial to the SIMS technique is severely compromised.
Molecular depth profiling achieved a quantitative status with the emergence of the C60 source
(34). With this probe, detailed studies aimed at elucidating the factors that lead to a successful depth profile were initiated. Trehalose thin films of several hundred nanometers were employed as a model (35). This species is attractive because smooth, uniform, and stable thin films are easily prepared in the laboratory via spin coating onto an Si substrate. Atomic force microscopy (AFM) is also a valuable tool, as it confirms film uniformity, detects the formation of topography that distorts thickness measurements, and reveals the physical shape of the erosion crater. Typical depth profiles are shown in Figure 4. Cheng and coworkers (36) utilized these data to create an erosion dynamics model that allowed the molecular sputtering yield, damage cross section, depth resolution, and altered layer thickness to be computed. They found that by using a 20-keV beam,
+
245 trehalose molecule equivalents per incident C60 molecules were removed, and interface widths of <10 nm could be measured. The model has been extended and made more sophisticated in the ensuing years (37).
- 34
- Winograd
319.11 56.94
[nm]
0.75 0.50 0.25
0.0
- 0
- 522.2089
[µm]
- 0
- 100
- 200
- 300
- 400
- 500
- 600
Depth, nm
Figure 4
Prototypical molecular depth profile of a 300-nm-thin film of trehalose spin coated onto an Si wafer using C60 for erosion. The red points are the quasimolecular ions of trehalose at m/z 325, [M−OH]+, whereas