Surface Chemical Analysis

Part of the Springer Handbooks book series (SPRINGERHAND)


The physical bases of surface chemical analysis techniques are described in the context of semiconductor analysis. Particular emphasis is placed on the SIMS (secondary ion mass spectrometry) technique, as this is one of the more useful tools for routine semiconductor characterization. The practical application of these methods is addressed in preference to describing the frontiers of current research.

Surface chemical analysis is a term that is applied to a range of analytical techniques that are used to determine the elements and molecules present in the outer layers of solid samples. In most cases, these techniques can also be used to probe the depth distributions of species below the outermost surface. In 1992 the International Standards Organisation (ISO) established a technical committee (TC) on surface chemical analysis (ISO TC 201) to harmonize methods and procedures in surface chemical analysis. ISO TC 201 has a number of subcommittees that deal with different surface chemical analytical techniques and this chapter will discuss the applications of these different methods, defined by ISO TC 201, in the context of semiconductor analyses. In particular, this discussion is intended to deal with practical issues concerning the application of surface chemical analysis to routine measurement rather than to the frontiers of current research. Standards relating to surface chemical analysis developed by the ISO TC201 committee can be found on the ISO TC201 web site (under standards development).

Traditional surface chemical analysis techniques include the electron spectroscopy-based methods Auger electron spectroscopy (AES or simply Auger) and x-ray photoelectron spectroscopy (XPS, once also known as ESCA – electron spectroscopy for chemical analysis), and the mass spectrometry method SIMS (secondary ion mass spectrometry). The ISO TC 201 committee also has a subcommittee that deals with glow discharge spectroscopies. Whilst these latter methods have been used more for bulk analysis than surface analysis, the information they produce comes from the surface of the sample as that surface moves into the sample, and so they have been finding applications in depth profiling studies.

One thing that is common to all of these surface chemical analysis techniques is that they are vacuum-based methods. In other words, the sample has to be loaded into a high or ultrahigh vacuum system for the analysis to be carried out. With the one exception of glow discharge optical emission spectroscopy (GDOES ), where the analysis relies upon the detection of photons, all of the techniques also depend upon the detection of charged particles. This requirement for vacuum operation necessarily imposes limits on the types and sizes of samples that can be analyzed, although of course instruments capable of handling semiconductor wafers do exist. The quality of the vacuum environment around the sample can also affect the quality of the analysis, especially with regard to the detection of elements that exist in the atmosphere around us. The size and complexity of surface chemical analysis equipment has arguably tended to limit the wider use of these powerful methods.

18.1 Electron Spectroscopy

In the electron spectroscopies, Auger and XPS, the surface of the sample is probed by an exciting beam which causes electrons to be ejected from the atoms in the sample. These electrons are collected and their energies analyzed. The two techniques are similar but subtly different.

18.1.1 Auger Electron Spectroscopy

In Auger, a beam of electrons is used to excite the sample. During the interaction of the primary electron beam with the sample atoms, core electrons are knocked out, creating vacancies in the inner electron shells. Electrons from outer shells can fall into the vacancy, thus leaving the atom in an unstable state and, in order to return to equilibrium, the excess energy the atom possesses can be dissipated in one of two ways: either by the emission of an x-ray photon with a characteristic energy (the basis of energy- or wavelength-dispersive x-ray analysis), or by the emission of a third electron (the Auger electron), with an energy determined by the difference in energy between the original core state and those of the two other levels involved. Clearly these energies are uniquely determined by the energy levels in the atom and thus provide a route for analysis. As three electrons are necessary for the Auger process to occur, AES is incapable of detecting hydrogen or helium, but all other elements produce characteristic Auger electrons. The energies of the Auger electrons can range from a few tens of electron volts to a few thousand electron volts. In this energy regime, electrons can only travel of the order of monolayers through a solid before an interaction occurs which causes a loss of energy, thus destroying the analytical information the electron possessed. It is this property which gives the Auger technique its surface sensitivity – although Auger electrons will be produced as far into the material as the primary beam can penetrate, only those produced in the top few monolayers can escape with their characteristic energy intact. The Auger electrons excited deeper into the sample lose energy before escaping and contribute to a background signal, as do the scattered primary electrons and the initial core electrons ejected at the start of the process. These scattered electrons produce a background signal against which the Auger electrons must be detected; this limits the analytical sensitivity that can be achieved by Auger. This is also why Auger spectra are sometimes displayed as the differential of the number of electrons against energy, because the small Auger peaks are more rapidly varying functions than the larger, slowly changing background signal and hence are enhanced by the differentiation process. Figure 18.1 shows an Auger spectrum from GaInAsP, where the spectrum plotted as the number of electrons as a function of energy is compared to the differential of the number of electrons as a function of energy.
Fig. 18.1

Auger spectrum from GaInAsP, shown as the number of electrons and the differential of the number of electrons as a function of energy

As the number of Auger electrons produced is proportional to the number of atoms excited, Auger offers the ability to perform quantitative as well as qualitative analyses, although some form of calibration is required, either through the use of local reference materials or instrument calibration and standard databases. So Auger offers quantitative analysis for all elements from lithium to uranium from layers only a few atoms in thickness. What gives Auger an extra dimension is the ability to profile into the sample by removing the outermost surface layers with an inert gas ion beam (usually argon) in order to expose the layers below. This sputter depth profiling is the inverse of the sputter deposition widely used to deposit thin layers of material. Argon is the most widely used ion beam in Auger depth profiling, as its effects are physical rather than chemical, although it should be noted that some chemicals can be modified by the sputtering process. For example, some metal oxides can be reduced by sputtering while others are not, and so any chemical state data inferred from atomic compositions in sputter depth profiles should be treated with caution. Physical effects can also occur, for example atomic mixing and the development of surface topography, which can distort the shape of buried features.

One of the great advantages of Auger is that, as the excitation is provided by electrons, the primary beam can be easily focused and scanned over the surface of the sample. By detecting the scattered electrons, a physical image of the sample is produced as in the scanning electron microscope, and maps of elemental distributions can be obtained by detecting the Auger signal as a function of beam position. These elemental maps can be time-consuming to acquire but can be useful when making a point or illustrating a book chapter; however, for practical analyses it is often adequate to identify the feature of interest from the physical image. Figure 18.2 shows aluminium and oxygen Auger maps from a contaminated bond pad. In the region of the contamination, the aluminium signal is reduced and the oxygen signal is higher compared with the uncontaminated regions.
Fig. 18.2

Optical image and Auger maps of a stained aluminium bond pad

In terms of hardware, Auger systems can be stand-alone systems comprising an electron beam column, an electron energy analyzer, an inert gas ion gun and sample handling stage and an associated vacuum chamber (or chambers), or form a part of a multitechnique system with x-ray sources for XPS analysis or a mass spectrometer for basic SIMS studies.

18.1.2 X-Ray Photoelectron Spectroscopy (XPS)

XPS is very similar to Auger in terms of the instrumentation and physics involved. The primary excitation, as the name implies, is, in this case, a beam of x-rays. The x-rays, often magnesium or aluminium Kα, eject core electrons from the surface atoms by the photoelectron effect. The kinetic energy of the emitted photoelectrons will be equal to the difference between the x-ray photon energy and the core level binding energy, and thus will be less than ≈ 1100 eV or ≈ 1400 eV for magnesium and aluminium Kα respectively. So, as with Auger, the mean free paths of the photoelectrons are of the order of monolayers in solid materials. As a core level vacancy is produced by the excitation, Auger electrons will also be present in the XPS spectra but, by convention, Auger spectroscopy refers to the electron-excited situation. Whereas in Auger electron spectroscopy the electron energy is usually referred to in terms of the electron’s kinetic energy, in XPS the electron binding energy is usually plotted as the ordinate in the spectra. This means that, whatever x-ray excitation is used, be it magnesium, aluminium or a more exotic material, the photoelectrons will appear at the same binding energy in the spectra but the apparent positions of the Auger peaks will change (on the binding energy scale) as their kinetic energy is independent of the excitation source. There are no scattered primary electrons (which are always present in Auger spectra) in the XPS spectra, so these spectra have better signal-to-background, and the photoelectron peaks are easily distinguished against the background arising from scattered photoelectrons produced deeper into the sample and other secondary processes.

As the problem of focusing electron beams is much simpler than focusing x-ray beams, XPS is perceived as a large-area technique, whereas Auger is the technique of choice for small area analysis. However, the relentless advances made in the performance and design of instruments means that XPS instruments can achieve spatial resolutions of the order of 1 to 10 μm. In XPS no charge is brought to the sample by the primary excitation, and so insulating samples are easier to analyze with XPS than with Auger; also, the photoelectron peaks show small energy differences in the peaks positions depending upon the local chemical environment of the atom from which they originated, the so-called chemical shift. While chemical shifts are present in some Auger peaks, this is the exception rather than the rule, and XPS is the technique of choice where information on the local chemical state of the surface is required. The physics of the XPS process is probably even better understood than the Auger process, and quantification of the spectra is relatively routine.

As with Auger, composition depth profiles can be produced by sputtering the surface of the sample with an inert gas ion beam. Again caution is advised when interpreting chemical state information from a surface that has been subject to ion bombardment. With both of the electron spectroscopies it is the surface of the sample that remains after sputtering that is analyzed in a depth profile, and there are two factors to be aware of (if not more). Once the passivating surface layer has been sputtered away, the surface of the sample may become chemically active and getter residual gas from the vacuum system. If the sample is a multicomponent material, one component may have a higher sputtering rate than the other, so as the sputtering process proceeds the surface will become depleted in the higher sputtering rate material. This process will continue until an equilibrium state is reached where the material leaving the surface is in the same ratio as the bulk composition; the corollary of this is that the surface will be enriched in the lower sputter rate material and so the composition of the material as measured by either XPS or Auger will be in error unless this effect is understood and accounted for.

Both Auger and XPS are capable of detecting all elements from lithium to uranium (and beyond), and have sensitivities in the parts per hundred to parts per thousand regime. The responses vary from element to element but typically sensitivities remain within an order of magnitude or so between elements.

18.2 Glow-Discharge Spectroscopies (GDOES and GDMS )

These are two apparently similar but quite unrelated techniques that rely upon glow discharges as the excitation source. In glow discharge optical emission spectroscopy (GDOES) a high-pressure glow discharge is used to sputter material from the surface of a sample, and this sputtered material is detected by the optical emissions it produces in the glow discharge. In glow discharge mass spectrometry (GDMS), a low pressure dc glow discharge is used to sputter material from the sample surface and ionized material from the discharge is extracted into a mass spectrometer for analysis.

GDOES is probably the simplest of all the surface chemical analysis techniques, at least as far as the vacuum requirements are concerned. There is no complex vacuum system, as needed for all of the other methods, and the sample itself sits with atmospheric pressure on one side of it while the opposite face acts as one electrode of a glow discharge cell that is pumped by a simple vacuum pump. A flow of high-purity argon gas flows through the cell, providing the sputtering and discharge gas and purging the cell of impurities and material removed from the sample. A window at the other end of the discharge cell transmits light from the discharge into an optical spectrometer. By using a spectrometer with a number of photomultiplier detectors, prepositioned at the known wavelengths of the expected element emission lines, data from a large number of elemental channels can be collected in parallel, making GDOES an extremely efficient analytical system. A scanning spectrometer can also be included in the instrument to provide a continuous spectral scan to detect emission lines from elements other than those built into the instrument, but this is, of course, a serial detection device and the advantages of parallel acquisition are lost.

Traditionally GDOES has been widely used for the analysis of metals and coatings on metals, but it is currently also finding application in the area of semiconductor materials. With the development of radio frequency glow discharge sources, the technique is capable of analyzing insulators, and with the high sputtering rates available, it can depth profile tens of μm into dielectric layers. With fast electronics and parallel detection, thin oxide layers can also be profiled. The sensitivity of the GDOES technique lies between that of Auger or XPS and that of SIMS. One potential weakness of the method is that the technique has no spatial resolution and the analysis area is millimeters in diameter. It is thus useful for large-area plain samples, but cannot be used with patterned material or to probe small features. The technique is useful for bulk analysis, but it is the ability to depth-profile into material, providing an insight into layer structures, that is its greatest appeal. However, because the technique has no ability to discriminate where the analytical signal is coming from, the quality of the depth profiles produced will be compromised by crater edge effects. In other words, while most of the analytical signal will originate from the bottom of the sputtered crater, there will always be some information that comes from the crater side wall. The consequence of this is that, with layered structures, layers closer to the surface will appear to tail into layers beneath them, even though the interface between the layers is abrupt. This effect can be seen in the depth profile shown in Fig. 18.3, which shows a GDOES profile into a dense wavelength division multiplexing (DWDM) structure.
Fig. 18.3

GDOES depth profile through a DWDM multilayer glass structure

Glow discharge mass spectrometry (GDMS) is a considerably more complex technique, at least from an instrumental point of view. Originally developed as a method of bulk analysis, GDMS is probably the most sensitive, in terms of the detection limit achievable, of all of the techniques being considered here. As with GDOES, in GDMS the sample forms one electrode in a simple glow discharge cell. However, in the case of GDMS, the discharge cell is mounted within a high-vacuum system. In its original form, the sample (typically be 1 mm2 by about 15 mm long) is placed in the center of a cylindrical cell into which argon is leaked at low pressure. By applying a dc voltage between the sample and the cell, an argon plasma is created which sputters the outside of the sample, removing material. This material, some of which is ionized but the majority of which is neutral as it leaves the surface, is ionized by a variety of processes as it passes through the glow discharge plasma. These ions are then accelerated into a high-resolution magnetic sector mass spectrometer where they are mass-analyzed and counted. Instruments can also be based on quadrupole mass spectrometers, but it is the magnetic sector instruments which offer the greater sensitivity. By sweeping the mass spectrometer through a range of masses, which can cover the entire periodic table, the major, minor and trace elements present in a sample can be determined. GDMS is a particularly powerful method of detecting the trace elements present in bulk semiconductor materials at levels down to parts per billion.

It is also possible to analyze flat, rather than matchstick-shaped, samples in GDMS. Just as in GDOES, the flat sample is positioned at the end of the discharge cell, and a cylindrical crater is etched into the sample surface. As with GDOES, with GDMS there is no spatial resolution, and the depth information from layered structures will be distorted by crater edge effects and loss of crater base flatness as it is not possible to discriminate between ions produced from the base of the crater and those produced from the sidewalls.

18.3 Secondary Ion Mass Spectrometry (SIMS)

SIMS is probably the most powerful and versatile of all of the surface analysis techniques and comes in the widest variety of instrumentations, from big, stand-alone instruments to bench-top instruments and add-ons to electron spectrometers. SIMS can offer chemical identification of submonolayer organic contamination, measurement of dopant concentrations, and can produce maps and depth profile distributions from nanometers to tens of μm in depth. However, no one instrument is going to be capable of all of these tasks, and even if it could it would not be able to achieve all of them at the same time.

SIMS, in its simplest form, requires an ion gun and a mass spectrometer. The sample is placed in a vacuum chamber and ions from the ion gun sputter the sample surface. Material is sputtered from the sample surface and some of this will be ionized, although in most cases the major part of the sputtered material will be in the form of a neutral species. The ionized component of the sputtered material is mass-analyzed with the mass spectrometer.

The technique has evolved in various directions from this common origin to produce a variety of subtly different variants of the SIMS technique, including dynamic SIMS (DSIMS ), static SIMS (SSIMS ) and time of flight SIMS (ToFSIMS ), each of which has its own distinct attributes. There are three main types of mass spectrometer used for SIMS analysis: the magnetic sector, the quadrupole and the time of flight, ToF. Dedicated depth-profiling SIMS machines, dynamic SIMS instruments, tend to employ either magnetic sector or quadrupole mass spectrometers. Magnetic sector instruments offer high transmission and high mass resolution capabilities, useful for separating adjacent mass peaks with a very small mass difference, for example 31P from 30SiH. Quadrupole mass spectrometers offer ultrahigh vacuum compatibility and, as well as being used in DSIMS instruments, smaller versions are also found as add-ons to Auger/XPS instruments and bench-top instruments. Time of flight instruments are remarkably efficient in their use of material in that the entire mass spectrum is sampled in parallel, whereas in the magnetic sector and quadrupole instruments the spectrum is produced by sequentially scanning through the mass range of interest. ToFs can also offer high mass resolution but profile relatively slowly because it is necessary to use a pulsed ion beam with a low duty cycle. Whilst static SIMS can be carried out on either a quadrupole or magnetic sector instrument, the ToF-based instruments are more suited to the task as less material is used in the course of the analysis. However, when dynamic SIMS, where only a few elements need to be monitored, is the goal, the magnetic sector or quadrupole instrument is the appropriate choice, although ToF machines can be used for shallow profiling.

Figure 18.4 shows a schematic diagram of the basis of these SIMS techniques. Ions from the ion gun bombard the sample surface and transfer momentum to the atoms in the sample creating a collision cascade. This cascade distributes the energy of the incoming ion amongst the atoms in the sample, causing them to be displaced from their original sites, possibly breaking some bonds between atoms whilst creating others. The bombarding ion also can become implanted into the target material, modifying the chemistry of the material in the process. None of this is peculiar to SIMS – it happens in any sputtering process, be it in Auger depth-profiling, GDMS analysis or sputter deposition systems. It is probable that some of the energy deposited in the sample will cause atoms or molecules to be ejected from the surface of the sample, and those atoms or molecules that leave the surface as ions can be collected and detected by the mass spectrometer.
Fig. 18.4

Schematic diagram of the SIMS process

In the early stages of sputtering, the atoms and molecules sputtered from the surface will originate from areas of virgin surface – in other words from sites that have not yet been damaged by the primary ion – and thus carry with them information about the chemistry of the outer molecular layers of the sample. As the sputtering process proceeds, the probability of the sputtered particles being emitted from an area that has been modified by earlier ion impacts increases. Thus, at the start of the process the sputtered particles are characteristic of the virgin surface, but they will eventually become characteristic of the ion beam-modified surface.

Static SIMS is concerned with the measurement of the sputtered molecules produced at the start of the process, where information about the surface chemistry of the sample can be obtained. In SSIMS a very low dose of primary ions is used, typically less than 1013 primary ions per square centimeter, so that there is a very low probability that the ions that are detected come from damaged material. Used in this way, SIMS can be used to identify organic contamination on surfaces and obtain information about the molecular structure of the sample. Figure 18.5 shows a typical SSIMS mass spectrum from poly(dimethylsilicone), a common surface contaminant, which can be recognized by prominent peaks in the mass spectrum at masses 73, 133, 147, 207 and 221, which originate from fragmentation of the parent molecule \(\mathrm{(CH_{3}{-}Si(CH_{3})_{2}{-}O{-}Si(CH_{3})_{2}{-}O{-}{\ldots}}\) \(\mathrm{{-}Si(CH_{3})_{2}{-}CH_{3})}\).
Fig. 18.5

Static SIMS spectrum from a thin film of poly(dimethylsilicone)

Whilst the abundance of molecular fragments that can be produced by the sputtering process is of great value for revealing chemical information about the sample surface in the SSIMS context, it can also be something of a problem when carrying out elemental analysis of trace impurities in simple matrices. For example, the spectrum shown in Fig. 18.6a, from an unknown Ga-based material, illustrates the number and complexity of molecular species produced by the SIMS process, which can make interpretation difficult. Fortunately, the distribution of the number of secondary ions with energy is different between atomic ions and molecular ions. The atomic ions, in general, have a broader energy distribution than the molecular ions, and its energy distribution tends to become sharper as the complexity of the molecular ion increases. For example, Fig. 18.7 shows the ion energy distributions of the Si+, Si 2 + and Si 3 + ions, illustrating the narrower energy distributions of the molecular ions. Thus, by selecting the ion energy range from which the spectrum is recorded, the molecular information can be suppressed, allowing the elements present in the material to be identified, see Fig. 18.6b.
Fig. 18.6a,b

SIMS mass spectrum from an unknown sample (a) without energy filtering, (b) with energy filtering

Fig. 18.7

The ion energy distributions of Si+, Si 2 + and Si 3 + ions

The intensities of the peaks in the SIMS mass spectra reveal more about the relative ionization probability and instrument transmission function than the number of species of that mass on the sample surface. The signals that are measured in SIMS are proportional to the numbers of ions produced, and this is not simply related to the amount of material present. The degree of ionization can vary enormously from element to element and matrix to matrix. For example, the numbers of positive ions formed by the inert gas bombardment of clean metal samples can be several orders of magnitude lower than the numbers produced from oxidized surfaces of the same metals under identical bombardment conditions. Clearly it is not the concentration of the metal atoms on the surface that is important, as there will be fewer metal atoms in the oxide layer than in the pure metal. It is the presence of oxygen that increases the ionization probability for material leaving the surface as positive ions.

In SSIMS, the chemical nature of the primary ion species is of relatively little consequence in terms of its effect on the number of ions produced by the surface, but in DSIMS, where it is the deliberately modified surface that is of interest, oxygen ion beams are widely used for analyses of electropositive species. Figure 18.8 shows how the silicon matrix signal from a silicon wafer with a native oxide layer behaves as the sputtering process proceeds. Initially there is a strong signal from silicon as the surface oxide layer is still present on the sample surface. The signal then falls as the oxide layer is sputtered away, and then slowly recovers in intensity as oxygen from the primary ion beam is implanted into the silicon surface. Eventually a steady state is reached where the material that is being sputtered is constant-composition oxygen-implanted silicon. The thickness of this transient region will depend upon the energy and angle of incidence of the primary oxygen ion beam. The higher the ion beam energy, the thicker the transient region. In the transient region the sputtering rate of the material and the ionization probability may well be changing, and precise quantification in this part of the profile may be difficult. This is of particular importance when the features of interest are very close to the surface and within the transient region. The technological importance of shallow structures has driven the development of very low energy primary ion beam columns to enable the characterization of shallow implants.
Fig. 18.8

The evolution of the Si+ signal from a clean silicon wafer as a function of sputtering time with an oxygen primary ion beam, showing the decay of the surface oxide signal and the build-up of the implanted oxide layer

Clearly, as is illustrated in Fig. 18.8 , the number of positive ions produced in the sputtering process depends upon the concentration of oxygen in the sample surface. In order to maximize the positive secondary ion yield, the sample surface needs to be saturated with oxygen, and this can be achieved with normal-incidence primary ion beams or by deliberately allowing oxygen to flood the sample surface as well as using an oxygen primary ion beam. In the example shown in Fig. 18.8, the primary ion beam was incident at approximately 45 and so complete oxidation of the silicon surface was not achieved, meaning that the silicon and oxygen signals in the equilibrium region are not as high as from the fully oxidized native oxide layer.

Not all species prefer to form positive ions in the sputtering process; some prefer to produce negative ions. Oxygen is a case in point, and indeed it would be difficult to conduct an analysis for oxygen using an oxygen beam. For electronegative species, it turns out that high negative secondary ion yields can be achieved if the surface is sputtered using cesium ions. An illustration of the relative ion yields of gallium and arsenic as positive and negative ions from a gallium arsenide surface with oxygen and cesium ion bombardment are shown in Fig. 18.9. Some species can be reluctant to produce either positive or negative secondary ions, but in some cases (such as zinc), good sensitivity can be achieved by using cesium ion bombardment and monitoring the cesium–element molecular ion, CsZn+ in the case of zinc.
Fig. 18.9a,b

Mass spectra from a GaAs wafer recorded (a) using Cs+ primary ion bombardment and negative secondary ion detection, and (b) using O 2 + primary ion bombardment and positive secondary ion detection

Despite the wide variations in secondary ion yields from element to element and matrix to matrix, the quantification of impurities in semiconductors can be relatively straightforward when reference materials are used. For example, Fig. 18.10a,ba shows a raw data depth profile of 11B ions implanted in silicon; knowing the areal dose of ions implanted into the sample and the depth of the crater sputtered into the sample, it is a relatively simple task (usually buried in the instrument software) to convert boron counts into concentrations (Fig. 18.10a,bb). By recording a signal from the matrix element, either during the profile as in Fig. 18.10a,ba, or in the crater after the profile, a relative sensitivity factor for the analyte (boron) in the matrix of interest (silicon) can be obtained which can then be used to determine the concentration of that analyte in another sample of the same matrix. Errors or uncertainties can arise from both the crater depth measurement and the implanted dose measurement. At the time of writing, there are only three metrologically traceable reference samples for SIMS (boron, arsenic and phosphorus in silicon) produced by NIST in the USA. For other species and other matrices, the analyst must rely upon locally produced materials with no traceability.
Fig. 18.10a,b

Reference sample of boron ion-implanted silicon, (a) raw data profile and (b) quantified depth profile

Care must be exercised when the analyte of interest is present in different matrices. For example, Fig. 18.11a-da–d show iron profiles, recorded under identical conditions (positive ion detection with O 2 + primary ion bombardment), from implants into InP, InAs, GaAs and GaP produced in the same implant run. While the relative sensitivity of iron to the phosphorus matrix signal is the same for InP and GaP matrices, there is a variation of the order of 20% in the arsenide matrices. The useful ion yield of iron (the total number of ions produced as a fraction of the number of atoms present) increases approximately four-fold in the gallium-based matrices compared to the indium-based matrices. The profile in the GaAs sample, Fig. 18.11a-d c, also shows a further complication: the arsenic matrix signal increases at about 1000 s into the profile. This could be interpreted as an upwards drift in the primary ion beam current, but it is actually caused by roughening of the GaAs surface as a result of the sputtering process. This effect is more common with oxygen primary ion bombardment than with cesium bombardment, and it has been shown that it is possible to reduce such effects by rotating the sample during analysis. Not only does the roughening modify the analytical signal, it also causes an apparent broadening of sharp features present in the sample as the surface roughness is convoluted with the feature width. The problems associated with sputter-induced roughness can become more severe when dealing with metal films but, once again, can be overcome by sample rotation.
Fig. 18.11a–d

Iron profiles, recorded under identical conditions, from implants into (a) InP, (b) InAs, (c) GaAs and (d) GaP, respectively

Another example of how relative sensitivity factors vary with the matrix is shown in Fig. 18.12a-c. Here a comparison is shown between the behavior of the Si and AsSi signals as the matrix is changed from GaAs to Ga0.16Al0.84As. Molecular signals such as AsSi are sometimes used to give increased signal compared to the atomic signal but, as in the example shown, the backgrounds are often similarly increased. What is interesting here is that whilst the ion yields from the Si and As signals increase, but at different rates with increasing Al content, the AsSi signals decrease. Ideally the analyst needs to be aware of, and able to take account of, such variations in response. Simply using the relative sensitivity factor derived for a GaAs matrix and applying this to high Al content material using a point-by-point normalization to the As signal can lead to a 30% overestimate of the silicon content. However, monitoring the AsSi signal instead and using the same normalization procedure can underestimate the silicon content by over 350%.
Fig. 18.12a–c

Depth profiles of silicon as Si and AsSi in (a) GaAs and (b) Ga0.16Al0.84As, and (c) variation in relative sensitivity factor as a function of aluminium content

With an ideal analytical technique the analytical signal will vary linearly with concentration over all concentrations. With SIMS, for uniform matrices, the measured signal is indeed linear with concentration for dilute systems. However, once the analyte concentration begins to exceed a few percent of the material, the signal may begin to vary in a nonlinear way with concentration. Thus, whilst SIMS is the ideal technique for quantifying minor and trace contaminants, it is less than ideal for determining matrix element concentrations. The most successful method of measuring elemental concentrations at high levels was achieved by monitoring the cesium molecular species (CsAl+ and CsGa+ and CsAs+) in GaAlAs. In GaAlAs the (\(\mathrm{CsAl^{+}/CsAs^{+}}\)) and (\(\mathrm{CsGa^{+}/CsAs^{+}}\)) ratios do vary linearly with the Al content.

Returning to practical analysis, another factor that needs to be considered in relation to DSIMS, and one that is related to the primary bombarding particles, is atomic mixing. The primary ions used to bombard the surface transfer momentum to the atoms in the sample surface. While some of the atoms will be sputtered from the surface, others will be driven deeper into the sample, thus broadening any concentration distributions that were present within the original sample. The higher the primary ion beam energy, the more severe the atomic mixing. Consequently, as the need to analyze shallower and shallower structures has increased, there has been a move to develop instruments that use lower bombardment energies and sources producing polyatomic ions. A simple example of how the analytical conditions can cause profile distortion is shown in Fig. 18.13a-c. These profiles were recorded from an arsenic-implanted and annealed polysilicon layer on a thin oxide on silicon. Using an oxygen primary ion beam striking the surface at oblique incidence with an impact energy of 5.5 keV (a) produces a profile that shows a spike followed by a rapid fall at the polysilicon/oxide interface. Using a cesium primary ion beam striking the surface at close to normal incidence with an impact energy of 14.5 keV (b) gives a much better detection limit with smaller interface spikes but with a much less abrupt fall in the arsenic signal. Profile (c), which uses the same bombardment conditions as profile (a) but combined with oxygen flooding of the sample surface (and consequently took longer to acquire), shows good agreement with profile (b) in terms of the size of the interface spike – which is exaggerated in profile (a) because of the increased ion yield from the oxide layer compared to the polysilicon – and good agreement with profile (a) in terms of the decay of the arsenic signal in the interface region, and has better sensitivity than profile (a) but less than profile (b). These profiles demonstrate the trade-off that the analyst must make when choosing analysis conditions.
Fig. 18.13a–c

Depth profiles of As in Si for different analytical conditions. (a) Oxygen ion bombardment, (b) cesium ion bombardment, (c) oxygen ion bombardment with oxygen flooding of the sample surface

18.4 Conclusion

The various surface chemical analysis techniques have their own strengths and weaknesses. No one method is suitable for all of the tasks the analyst faces; sometime one technique is sufficient to address the problem at hand, sometimes a combination of them is required. However, the approach should be successful if the technique(s) is (are) fit for the purpose of the task.

Copyright information

© Springer International Publishing AG 2017

Authors and Affiliations

  1. 1.Loughborough Surface Analysis LtdLoughboroughUK

Personalised recommendations