ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 19 Surface Analysis in an industrial Environment Robert K. Wild Nuclear Electric plc, Berkeley Technology Centre, Berkeley, Gloucestershire GL 13 9PB Introduction Industry uses surface analytical techniques to ensure quality, to improve productivity, to solve problems and to increase safety. A high-pressure steam pipe may have failed causing considerable damage, possible risk to staff and perhaps the public, and loss of production. Surface analytical techniques can be used to aid the investigation into the cause of the failure and could subsequently be used to monitor the quality of components to be used in future to prevent the recurrence of such a failure. Quality control is particularly important in the semiconductor industry where the surface condition of the component is of vital importance to the performance of the resulting electronic device.There are many surface analytical techniques available at present and decisions must be made as to which is the most suitable for the particular application. This is more important in surface studies because these techniques tend to be expensive and very few industrial companies are large enough to run a department which operates all the possible techniques. Increasingly, use is being made of the facilities and expertise that exist in many universities and much of the specialized analytical work is being undertaken under contract. As a result the relationship between industry and the universities has become much closer in recent years. This paper attempts to identify the different surface techniques, illustrate their strong and weak points and give some examples of where such techniques can be useful. For a complete description of the techniques available the reader is referred to Rivikre.1 Surface Analytical Equipment When deciding which technique is the most suitable to solve a particular problem or which piece of equipment should be purchased by a laboratory, careful consideration needs to be given to the requirements of the job and these then need to be matched to the techniques and equipment.If it is important to be able to obtain a good image of the surface and the analytical information is required from small areas then good spatial resolution is essential; if high sensitivity to certain elements is required or the chemical state of the surface must be determined then a different technique may be necessary.Some of the major techniques available in commercial form are listed in Table 1. In this table the type of probe, the species detected and the process involved in the technique are given. In any investigation it is important to cause as little damage, to the surface that is being probed, as possible. For this reason the least damaging technique should be considered first. In general, photons cause less damage than electrons which are, in turn, less damaging than ions or atoms. The energy and intensity of the beam must also be considered, an intense photon beam such as encountered in some lasers can be far more damaging than a low intensity ion beam, while a photon with a wavelength of a few hundred nanometres would cause less surface perturbation than an X-ray photon with a wavelength of a fraction of a nanometre.Infrared spectro- scopy (IRS), laser Raman spectroscopy (LRS) and X-ray photoelectron spectroscopy (XPS) all probe with photons and can identify elements and give some chemical state informa- tion. They do, however, have the drawback that the spatial resolution is limited to approximately 1 pm for the IRS and LRS techniques and to approximately 10 pm for XPS. There is also the difficulty of ensuring that the surface is being analysed. Both visible and infrared radiation penetrate several hundred nanometres into most solids and even when using reflection infrared, sampling is from a relatively large volume.If spatial resolution is an important criterion then electron or ion probes must be employed. Both Auger electron spectro- scopy (AES) and secondary ion mass spectrometry (SIMS) can give spatial resolutions of a few tens of nanometres while at the same time giving good element detection with some Table 1 Surface techniques available to industry Technique Infrared Raman (laser) X-ray photoelectron spectroscopy Low energy electron diffraction (LEED) Auger electron spectroscopy (AES) Secondary ion mass spectrometry (SIMS) Sputtered neutral mass spectrometry (SNMS) Scanning tunnelling microscopy (XPS) (STM) Probe Photon Photon Photon Electron Electron Ion Ion High field Detected Photon Photon Electron Electron Electron Ion Atom Electron current Process Bond vibrations Electronic excitation Ionization Elastic scattering Auger Sputtering Sputtering Electron tunnelling Table 2 Relative merits of surface analytical techniques Element Technique detected Sensitivity Quantitative Infrared Raman (Laser) XPS LEED AES SIMS SNMS STM Most Most All None All >He All All Some Few % Few YO 1 Yo N/A 1 Yo PP" NIA PPb Semi Semi Yes NIA Yes No Yes NIA Spatial resolution ym ym 10 ym <1 ym 20 nm 50 nm 50 nm 0.1 nm Chemical state Yes Yes Yes No Some Yes Yes Some Structural information No No No Yes No No No Yes NIA = Not applicable.20 ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 Fig.1 Failure of a weld in a power station steam chest chemical state information. The relatively new technique of scanning tunnelling microscopy (STM) and the atomic force microscope (AFM) are, in principle, non-damaging and can give better resolution than any of the other techniques.In practice, considerable expertise is required to obtain the best from these instruments in a non-destructive manner. The STM can also give valuable structural information which together with low energy electron diffraction (LEED) can be useful in determining defect concentrations and surface damage. Finally, if sensitivity is important then the ion bombardment techniques should be employed. Most surface analytical techniques have a detection limit in the region of 0.1 at.-% but SIMS and sputtered neutral mass spectrometry (SNMS) can achieve ppm and in some cases ppb detection levels, often with high spatial resolution.The relative merits of the various techniques are listed in Table 2. Here the elements detected, the sensitivity, quantita- tive analysis, spatial resolution, chemical state and structural information that can be obtained are included. Applications Industry has a wide range of needs for surface analytical techniques. These range from control and development of semiconductors to mechanical properties of superalloys exposed at high temperature to large doses of radiation, from corrosion in both gaseous and aqueous environments to the performance of new materials such as carbon-fibre composites and metal-matrix composites and from catalysis in the chemical industry to adhesion, or the lack of it, between materials. Attempts will be made to give some examples of the use of surface techniques in a few of these industrial applications.The examples will come mainly from the nuclear or power generating industries but the principles apply to a much wider range of industries. North HP steam chest Fig. 2 Schematic diagram showing location of a steam chest failureANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 21 7 , I 1 t Crack tip Fig. 3 SAM Fracture surface of the crack from the steam chest opened in Mechanical Properties The mechanical properties of steels are frequently determined by the segregation of trace elements to grain boundaries together with the build-up of particles, typically carbides, at these boundaries. An example where a steel component has failed in a catastrophic manner is shown in Fig. 1. This is a steam chest in a conventional power station and is a relatively large component used to transport high-pressure steam between the boiler and the turbines and is shown schematic- ally in Fig.2. It is constructed from a ferritic steel and has failed in the vicinity of a weld. Here the history of the failure can be determined from crack dating but to find the cause of the failure we must look to the state of the bulk metal. Ferritic steel can be induced to fracture in an intergranular manner at liquid nitrogen temperatures. It is, therefore, a relatively straightforward matter to produce a fracture surface for analysis in a scanning Auger microscopy (SAM) or SIMS instrument. The specimen has to be cooled to close to 77 K and a sharp impact applied to the specimen. A typical fracture surface is shown in Fig.3. This was obtained by cutting a specimen in such a way that the pre-existing crack could be used as a notch to initiate the fracture path. The grains, ahead of the crack tip, are clearly visible as smooth surfaces with sharp boundaries between adjacent grains and the tip of the crack can also be seen running across the image. This problem requires a technique with good spatial resolution to identify 6 5 - w X Z F m F L3 c w 4 1 I x 3 2 1 n 100 200 300 400 500 600 700 800 900 1 ~ 0 0 Kinetic energyIeV Fig. 4 Auger spectrum recorded from the fracture surface in Fig. 3 Fig. 5 Example of intergranular fracture in alloy PE16 elements on the grain boundary. Auger spectroscopy has been used but SIMS could have been used if very high sensitivity were required.An Auger spectrum recorded from a grain is shown in Fig. 4. In this instance two embrittling elements, tin and copper, are observed to be present on the grain surface and these were undoubtedly, at least in part, responsible for the failure of this component. Phosphorus is a much more common segregant that causes embrittlement in ferritic steels and has resulted in the cracking of countless bolts in industrial plants. Consideration is being given to the use of this technique to determine the condition of components so that those that are susceptible to failure could be replaced and research is underway to predict more accurately the failure of steels from a knowledge of phosphorus content and grain size. The mechanical properties of nickel-based alloys are also determined by grain boundary segregants but these alloys are ductile at liquid nitrogen temperatures and it is necessary to hydrogen charge the specimen and then fracture by using a slow tensile strain.A nickel-based alloy PE16 (44 + 34 + 16% m/m, Ni + Fe + Cr, respectively) is used to support the heavy fuel assembly in our advanced gas cooled reactors (AGR) and it is important that the integrity of this rod is guaranteed throughout its life. The alloy becomes radioactive during its life yet must, of course, be examined. We have, therefore, developed a tensile fracture stage that is capable of fracturing small active specimens in the form of matchsticks less than 8 X 1 X 1 mm to reduce the operator dose.* The fracture surface of one of these alloys is reproduced in Fig.5 and looks similar to the ferritic impact surface. The Auger spectra from a grain boundary in a specimen in the ‘as manufactured’ condition is compared with the spectrum from a specimen following22 r: 0 .- .I.- F al 2 25 0 0 4- ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 - 7 (a) 6 - I I tNi I - Fe Fe 7- Ni I A 0 100 200 300 400 500 600 700 800 900 1000 - Z - 7 ] ( b ) 6 t "4 0 100 200 300 400 500 600 700 800 900 1000 Kinetic energylev Fig. 6 Auger s ectra from grain boundary of ( a ) as manufactured alloy PE16 and 8) irradiated alloy PE16 irradiation (Fig. 6). It is clear that there has been an enrichment of nickel and a depletion of chromium and iron together with segregation of phosphorus and silicon. The nickel enrichment is considered to be the result of the diffusion rates of the major alloying species.Irradiation produces a large number of point defects (vacancies and interstitials). These move through the lattice in a random manner but on encountering a sink such as a grain boundary will be annihilated. This produces a net flux of point defects towards the grain boundaries. A flow of vacancies towards grain boundaries implies a flow of atoms in the opposite direction; however, the rate of flow will depend on the jump frequency of the particular atom.3 It has been determined that the jump frequency, J , of chromium > J(iron) > J(nickel);4 hence, as chromium and iron flow away from the boundary at a greater rate than nickel the boundaries become enriched in nickel by default.The effect can be reversed by annealing the alloy at 600 "C for 1 h. Fig. 7(a) and ( b ) show the levels of grain boundary components for the major and trace elements, respectively, in the 'as received', irradiated and annealed conditions. This indicates that the annealing returns the alloy at least partially to its original condition. Corrosion In the nuclear industry the 20% Cr-25% Ni/Nb stabilized stainless steel is used to clad the uranium dioxide fuel and must operate at temperatures of up to 1000 K in a C02-1% CO gas environment at a pressure of 40 atm (4.05 MPa). A potential problem was envisaged with this alloy. During thermal cycling it was thought that the oxide which formed on this alloy might J. o t 1 1 I I Arch Matrix Arch G.B. lrrad Anneal Treatment 4 3 s 1 C 0 .- c c F 2 al Z ' 0 0 -1 I I I I Anneal Arch Matrix Arch G.B.lrrad Treatment Fig. 7 Grain boundary composition of alloy PE16 as a function of various treatments. (a) For bulk elements: A, nickel; B, iron; and C, chromium. ( b ) For trace elements: A, silicon, B, phosphorus; and C , molybdenum spa11 off and be transported around the coolant circuit. The spalled oxide would be radioactive and could lead to high levels of radiation at points around the coolant circuit. It is, therefore, important that the mechanisms of corrosion and spallation are understood and that the composition of the oxide is determined. Several techniques can be utilized to determine the oxide composition and to determine the mode of spallation. The XPS, AES, SIMS, SNMS, Raman and laser Raman tech- niques can be used to characterize the outer oxide surface.Laser Raman spectroscopy has been applied to the outer atom layers of the oxide that forms on this alloy. Fig. 8 shows a series of Raman spectra obtained from oxide films formed on 20% Cr-25% Ni/Nb stabilized stainless steel following 24 h oxidation in C02-4% CO-300 ppm v/v CH4-300 ppm v/v H20-400 ppm v/v H2 at temperatures between 600 and 900 "C.s At the lower temperature only the spinel oxide forms but as the temperature increases the rhombohedra1 Cr203 phase begins to form. To characterize the oxide the composition must be obtained as a function of depth. This can be carried out by depth profiling through the oxide using ion bombardment while using a surface-sensitive technique to monitor the composi- tion.As spatial resolution was not important and the chemical state of the oxide was required XPS was used in combination with argon ion bombardment. The results of this are shown in Fig. 9 and indicate that the oxide is duplex in nature with an outer layer approximately 0.4 pm thick consisting of a chromium manganese layer on top of a chromium-rich layer also 0.4 pm thick. As a large area has been depth profiled, X-ray diffraction (XRD) could be used to obtain structural information concerning the oxide layers. X-ray diffractionANALYTICAL PROCEEDINGS. JANUARY 1993, VOL 30 23 traces taken before depth profiling and after 0.4 pm has been removed are shown in Fig. 10. The spectrum recorded before etching showed both spinel and rhombohedral oxide peaks but the diffraction pattern recorded after profiling to a depth of II I /I Fe304/spinel Spine' (broad feature) c 4- - 1000 900 800 700 600 500 400 300 Raman shiftkm-' Fig.8 Raman spectra obtained from oxide film formed on 20% Cr-25% Ni/Nb stabilized stainless steel for 24 h in C 0 2 at various temperatures Oxygen Ox id e/oxid e Oxide/metal Oxide'gas interface - interface i n te rface , . . . ...- 0 0.2 0.4 0.6 0.8 1 .o 1.4 Depth profiled/pm Fig. 9 Cr-25% Ni/Nb stabilized stainless steel XPS depth profile through the oxide formed on a 20% 0.4 pm showed only rhombohedral peaks, indicating that the outer layer was the spinel oxide. The bulk of the oxide could therefore, be characterized as an outer chromium-manganese spinel superimposed on a rhombohedral chromium oxide.If one is attempting to determine very low levels of certain trace elements then SIMS should be used. A low level of yttrium was added to an experimental alloy to determine if it could usefully be used to reduce the rate of corrosion and SIMS was then used to determine the position of the yttrium in the oxide. bSpinel+Cr203e- 20-25/Nb Steel 106 105 v) 104 103 F I v) c 3 c1 .- v) S g 102 - . - 0 10 20 Sputtering time/min 30 Fig. 11 20% Cr-25%0 Ni/Nb stabilized stainless steel SIMS depth profile through oxide formed on yttrium doped Stage / I l l 1. Vapour deposit nickel foil at -300 "C in glow cut -,-A discharge Stage 3. Cut along end of nickel dated foil stresses between oxide and nickel cause nickel Cut - to peel away exposing metal/oxide Nickel ' Oxide Foil interface Stage 2.Cut along edges of nickel plated foil Fig. 12 Sputter ion placing method to reveal the rnetal/oxide interface 65 60 55 50 45 40 35 30 25 2H Fig. 10 XRD Datterns recorded from the surface of the oxide in Fig. 8:Ta) before de'pth profiling and ( b ) after profiling through 0.4 pm. 8, austenite: R, rhombohedral phase (= Fe203 in Crz03): S, manganese spinel (MnFe, Cr2-, 0,: 0 d x d 2) Cr-25% Ni/Nb stabilized stainless steel Fig. 13 ( a ) Metal and ( b ) oxide sides of the interface on 20%24 ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 Fig. 14 Auger element maps from the metal side of the metalloxide interface: ( a ) SEI; ( b ) Fe; ( c ) Cr; and (d) Si An example of a SIMS depth profile through an oxide on this steel that had been implanted with yttrium is reproduced in Fig.11.6 The level of yttrium can be seen to vary by more than four orders of magnitude across the oxide layer. An attempt was made to simulate spalling of the oxide layer using a technique of sputter ion coating (Fig. 12).7 Nickel was plated onto the outer oxide surface in the presence of an argon ion plasma discharge at a temperature of 500 K. The plasma discharge cleans the outer surface and aids the bonding of the nickel to the oxide while the elevated temperature causes strains to be set up on cooling to room temperature between the layer and the oxide which in turn causes the oxide to spa11 away from the substrate alloy. Scanning Auger microscopy was used to analyse the two sides of the metal/oxide interface. Secondary electron images of the interface are shown in Fig.13.8 The oxide appears to have pulled away from the metal pulling out plugs of oxide from the positions of the grain boundaries. The spatial resolution of SAM is of the order of 100 nm and is therefore more than adequate to study boundaries of 10 pm in size. Spectra can be recorded from various points on the grain boundary surface but it is qualitatively more instructive to produce element maps across the surface. Typical examples are shown in Fig. 14 for the metal side of the interface for chromium, silicon and iron. It can be seen that chromium and silicon concentrate at the grain boundaries on the metal side. Similar studies on the oxide side of the interface indicate that chromium and silicon concen- trate at the grain boundaries but that silicon alone is present at the grain centre.Depth profiling through this silicon layer at the centre of the grain shows that the silicon layer is only 20 nm thick. Laser Raman spectroscopy can be used to study the change in stresses at points across the oxide and this may be important in determining the adherence of the oxide during thermal cycling. Stress in solids affects the molecular vibra- tions, generally moving them to higher frequency and thus producing a Raman shift to higher frequencies. This change in Raman frequency with stress has been used to determine the stress across a Cr203 oxide formed on stainless steel.9 Fig. 15 shows the appearance of the corrosion scale with low ridges formed over the grain boundaries of the substrate metal.The stress-induced shift in the Raman band of Cr203 is plotted against position in Fig. 15 and reveals significant stress relief at the grain boundaries. 560 ""I 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Dista nce/ym Fig. 15 laser Raman Determination of stress in oxide using frequency variation in Combining these results allows the oxide to be charac- terized and the mode of spallation to be determined. Indeed the oxide spalls by fracturing on the metal side of the silicon oxide layer at the grain centres but it fractures through the oxide at the grain boundaries. This has the effect of leaving a plug of silicon-chromium oxide in the grain boundary whichANALYTICAL PROCEEDINGS, JANUARY 1993. VOL 30 25 . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . ' oxide Outer) Spinel Carbon rich layer Silicon rich layer ri::} Rhombohedral Fig. 16 Schematic diagram of oxide on 20y0 Cr-25y0 Ni/Nb stabilized stainless steel Fig. 17 Low energy diffraction pattern from silicon(ll1) surface reduces cation diffusion along the boundary and hence reduces the subsequent rate of oxidation. This is summarized in Fig. 16 where the oxide composition is shown in schematic form and the spallation path indicated. Semiconductors Semiconductor production and development have become fundamental to the running of the world as we now know it. Almost all electronic equipment contains many of these components. The level of reliability must be high and to achieve this, manufacture must be carried out in conditions of maximum cleanliness to prevent surface contamination.Sur- face analytical techniques are used extensively to monitor cleanliness, to determine dopant concentrations and to aid in developing new products. The quality of the silicon can be determined in a number of ways, the bulk structure can be determined using conventional XRD techniques but the structure of the surface requires surface diffraction and here LEED can identify the structure. In one example the LEED pattern from an Si(ll1) surface is reproduced in Fig. 17. This diffraction pattern might imply that the surface structure was perfect as it appears to give only reflections in the positions expected from a good silicon crystal. However, if the same surface is examined using STM it becomes clear that the surface is far from perfect and in fact contains many defects (Fig.18). 10 Analysis of the material can be carried out using a technique with spatial resolution better than 1 um as this will allow the various pathways in the device to be resolved. This makes XPS a borderline technique although some good work has been carried out using the imaging ESCASCOPE, but AES and SIMS are most frequently used. The former is generally preferred if accurate quantitative determinations are required but SIMS must be used to detect low levels of impurities and dopants. Frequently, several layers of material are deposited onto a silicon substrate and these may be only a few tens of nanometres thick and it is important to be able to measure the thickness of the layers following deposition.Depth profiling techniques must be used, either in conjunction with AES or SIMS. Fig. 19 shows a depth profile through an Fig. 18 Scanning tunnelling microscope image of silicon( 11 1) surface AlGaAs multilayer quantum well structure. It contains 30 repeating layers of GaAs substrate each 14 nm thick with 6 nm thick layers of A1,Gal-,As between them all, on top of a 400 nm thick layer of GaAs on 9 layers of 4 nm GaAs, between 4 nm thick A1,.Gal-,.As, on GaAs substrate. This demonstrates how many very thin layers can be successfully resolved. Many semiconductor devices are encapsulated in plastics which presents the analyst with particular problems, particularly if he/she is attempting to identify a cause of failure. In these cases the non-destructive technique known as infrared micro- scopy (IRM) can be utilized.This technique can be used to identify failures in encapsulated devices." The trick is to mount the complete device and then polish back mechanically to the silicon substrate (Fig. 20). As silicon is transparent to infrared radiation and IRM has a spatial resolution of 1 pm the surface of the device may be investigated through the silicon substrate. In particular, corrosion at pin locations can be detected and the effect of operating the device can be monitored using this technique. In the example shown here a 14 pin CMOS chip is studied. Fig. 21 shows this component26 0 50 100 150 200 250 300 350 400 450 500 Sputtering time/mi n Fig. 19 Depth profile through an MBE grown AlGaAs multilayer quantum well structure Fig.20 operating surface Polishing a CMOS device to allow infrared studies of the ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 (4 after being polished to the back side of the silicon substrate and viewed by illuminating with infrared radiation. By connecting the device and applying a voltage to the approp- riate pin the device could be triggered to operate. By raising the voltage further it was possible to increase the dissipation in the polysilicon input resistor until the associated thermal emission was clearly identifiable as the light coloured region in Fig. 21(b). Continued application of such power, for even short periods of time, would be likely to result in permanent damage to the device. In this study, fusion from the input resistor occurred after approximately 1 s as can be seen in Fig.21(c). Clearly this method of viewing an interface from below has considerable potential and may be applicable to IR spectroscopy. Conclusions Surface analytical techniques are being used increasingly by industry to solve problems and maintain quality. This has become possible as a result of the basic work undertaken at universities and other major research establishments to understand the techniques fully and as a result of manufactur- ers producing reliable, easy to use equipment that can give results for a large number of specimens quickly. Some techniques are used less than they might be by industry because they have not yet reached this level of sophistication. Laser Raman instruments have only recently been developed to the stage where they are sufficiently robust to be used by industry while STM is a technique with potential that is still to be realized. These and some other techniques will remain in universities for some years yet, but will be utilized by industry Fig. 21 Infrared microscopy of a 14 pin CMOS chip: ( a ) after polishing: ( b ) during operation: and (c) after operation to a level which caused failure on a contractural basis thus maintaining the important link between the two areas. References 1 Riviere. J. C.. Surface Analytical Techniques, Clarendon Press, Oxford, 1990.ANALYTICAL PROCEEDINGS, JANUARY 1993, VOL 30 27 2 3 Nettleship, D. J.. and Wild, R. K., Surface Interface Anal., 1990, 16. 552. Perks, J. M., Marwick, A. D., and English, C. A., Proc. Radiation Induced Sensitisation of Stainless Steels, Berkeley Nuclear Laboratories, Berkeley, 1986, pp. 86-98. Marwick, A. D., Piller, R. C., and Horton, M. E., AERE Report 10895, Atomic Energy Research Establishment, Har- well, 1983. Bennett, M. J., Proceedings of the Conference on UK Corro- sion, 1984, p. 43. 4 5 6 Bennet, M. J., Bishop, H. E . , Chalker, P. R., and Tuson, A. T., United Kingdom Atomic Energy Authority Report A E R E R 12540, Atomic Energy Research Establishment, Harwell, 1987. 7 Coad, J . P., and Wild, R. K., Appl. Surface Sci., 1983,14,321. 8 Wild, R. K., Spectrochim. Actu, Part B , 1985, 40, 827. 9 Gardiner, D., and Bowden, M., Microsc. Anal., 1990, 20, 27. 10 Tear, S. P., Microsc. Anal., 1990, 19, 7 . 11 Ford, D. J., Microsc. Anal., 1990, 17. 39.