Outsourcing of Analysis: Choosing the Right Technique

Scott Baumann Evans Analytical Group / Evans Texas

Introduction Over the past 25 years the availability of high-quality commercial analytical services has grown significantly. Now, many quality outsourcing options are available to failure analysts and engineers. Not only are duplicate capabilities often available for times when in-house labs cannot keep up with the work flow due to large numbers of samples or instrument failure, but state-of-the-art instrumentation and expertise are also available to even the smallest companies. In some cases, failure analysis personnel may only be familiar with the available in-house techniques and may think that all analysis options for a certain problem have been exhausted, when usage of outsourced analytical techniques can actually provide new insight and answers into 'unsolvable' problems. This paper will summarize some of the state-of-the-art analysis techniques that are commercially available, and explain the critical factors for choosing the appropriate technique. In addition, some 'tricks of the trade' will be offered with respect to improving the likelihood of success, expediting analyses, and improving cost effectiveness.

Why Outsource? Some reasons for outsourcing analytical work are: 1) The needed analytical technique is not available

in-house 2) An external lab has special expertise in a certain

type of analysis, or specialized instrumentation that affords better detection sensitivity or spatial resolution versus what is available in-house

3) The in-house lab is backlogged with work 4) Emergency situations when fast turnaround of

results are crucial and the in-house lab may not be able to analyze samples fast enough.

5) An unbiased, third party is needed to resolve a dispute, or validate unexpected or unusual results.

Another reason for outsourcing is significant cost savings. In particular, the cost for state-of-the-art instrumentation and capable analytical personnel can be quite large. Outsourcing of the work can provide significant cost savings compared to purchasing and staffing an instrument in-house for a particular type of analysis that is needed sporadically, e.g., only

during startup or transfer of a process. In addition, there are no fixed costs associated with outsourcing, so if business levels drop, or if the need for a particular type of analysis goes away, so do the costs of that technique. Because larger commercial laboratories are combining work from numerous sources, sufficient work exists to support the purchase of state-of-the-art instrumentation. The quality of outsourced results can then be significantly better than work performed in-house on an older instrument. In many cases, at commercial labs there is sufficient work to justify dedicating an instrument to a particular analysis or modifying the instrument to provide better performance for that specific analysis. For example, SIMS instruments can be modified with specialized pumping and be dedicated to the analysis of atmospheric contaminants in materials. Because only certain types of samples are analyzed on these instruments, significant improvements in detection limits for atmospheric contaminants can be obtained compared to an in-house instrument that has to be used for many different types of analysis. There can be disadvantages to outsourcing that should be considered: 1) The scientist analyzing the samples may not be

familiar with the specific devices or processes being analyzed. This is normally not a problem, as long as sufficient information is provided. However, an in-house expert who has analyzed hundreds or perhaps thousands of a certain device would most likely have unequalled and extensive knowledge about product evolution and problem history.

2) The sample will have to be transported from your facility to the external lab, either by shipping or being hand-carried. Depending upon the location of the external lab this can result in longer response times for an emergency analysis than an in-house lab.

3) Justification of spending for services provided by an external lab may be difficult if in-house lab services are considered to be “free” resources or capable to handle any problem.

Labs that provide outsourcing services are aware of these obstacles to sending out samples and have taken steps to reduce the impact of these issues.

Understanding the Physics of

Different Analytical Techniques Over the past 30 to 40 years a wide range of surface and microanalytical techniques have evolved. Each technique has its own unique capabilities that are related to the particular physical interaction involved with that technique. Table 1 lists some of the commonly used analytical techniques and their acronyms, while Figure 1 illustrates schematic

diagrams of the physical process involved in some of these analyses. With the exception of SPM/AFM, all of the techniques involve the interaction of some type of particle (electron, ion, or photon) with the sample material. The physics of each particular interaction affect the limits of lateral resolution, depth resolution, and detection sensitivity for each technique. Understanding these interactions, and more importantly the limitations they impose on a technique, can be crucial when selecting an analytical technique.

Figure 1 Physics of surface analysis techniques

AAS Atomic Absorption SpectroscopyAES Auger Electron SpectroscopyFE AES Field Emiss ion Auger Electron SpectroscopyEELS Electron Energy Loss Spectroscopy FTIR Micro Fourier Transform Infrared SpectroscopyGDMS Glow Discharge Mass SpectrometryICP-MS Inductively Coupled Plasma Mass Spectrometry LA-ICPMS Laser Ablation Inductively Coupled Plasma Mass SpectrometryRaman Raman SpectroscopyRBS/HFS Rutherford Backscattering Spectrometry/Hydrogen Forward Scattering SEM-EDS Scanning Electron Microscopy-Energy Dispers ive X-Ray SpectrometryFE SEM Field Emiss ion SEMSIMS Secondary Ion Mass SpectrometrySPM/AFM Scanning Probe Microscopy/Atomic Force Microscopy TEM Transmiss ion Electron MicroscopyTOF-SIMS Time-of-Flight Secondary Ion Mass SpectrometryTXRF Total Reflection X-Ray FluorescenceVPD Vapor Phase Decompos itionXPS/ESCA X-Ray Photoelectron Spectroscopy/Electron Spectroscopy for Chemical Analys isXRD X-Ray DiffractionXRF X-Ray Fluorescence

Table 1 Acronyms for commonly used analytical techniques

TXRF

For example, electron beams can be focused to very small spots. This allows electron-beam based techniques such as SEM, TEM, and AES to provide much higher spatial resolution than techniques such as XPS or FTIR, which use photons to probe a sample, or SIMS and RBS, which utilize ions[NU1]. While X-ray and ion-beam based techniques may lack the spatial resolution of electron beam techniques, they offer other unique advantages such as sensitivity to chemical state (XPS), organic structure (FTIR, Raman, TOF-SIMS), or absolute concentration sensitivity (Dynamic SIMS). Shown in Figure 2 is a chart plotting detection sensitivity versus analytical spot size for a variety of analytical techniques. This chart, along with Table 2, which summarizes crucial parameters for a number of techniques, allow the reader to begin the process of selecting the appropriate technique for a given problem. One feature in Figure 2 of particular note is the “physical limit” lines shown in the lower left corner of the chart. These lines show the point at which a technique would be sampling a single atom of a given impurity. For example, AES samples to a depth of ~30Å, so there are about 2x106 total atoms available for AES analysis in a silicon device area that is 1,000Å x 1000Å in size. If an impurity were present at a concentration of 0.5ppm in this material,

then there would be 1 atom of that impurity available for AES to detect in the described volume. The best that can be achieved is to detect every available atom. In reality, most techniques cannot even approach this level of sensitivity. While useful, it is not essential to understand the physics involved in a particular technique before outsourcing an analysis. It is important, however, that the scientist performing the analysis has a good understanding of the physics involved of the technique being used and the implications when applied to a particular sample. For example, ion beam mixing effects will affect the depth resolution for SIMS analysis, and the relationship between electron energy and escape depth will affect the sampling depth of AES analysis for different elements. Normally commercial analytical labs have experienced and well trained staff, however, exceptions can occur. Also, many universities offer analytical services with mixed levels of experience of the personnel performing the analyses, e.g., from highly experienced staff to inexperienced graduate or undergraduate students. Inquiries regarding the experience level of staff, how a technique is performed, or the significance of analytical results are appropriate.

Figure 2 Detection sensitivity versus analytical spot size

Table 2 Analytical Technique Summary

Analytical Technique Typical Applications Signal Detected Elements

Detected Organic

Information Detection

Limits Depth

Resolution Imaging/ Mapping

Lateral Resolution

(Probe Size)

AFM/SPM Surface imaging with near atomic resolution Atomic scale roughness - - - 0.01 nm Yes 1.5 - 5 nm

FE Auger Elemental surface analysis, microanalysis, micro-area

depth profiling

Auger electrons from near-surface atoms Li - U - 0.1 - 1 at% 2 - 6 nm Yes <15 nm

FIB

Cross sections - especially Cu, resist, defects Thin sections for

STEM/TEM

Secondary and backscattered electrons

Secondary ions

B - U (EDS mode) - 0.1 - 1 at% Yes

>3 nm with SEM

> 7 nm with FIB

FTIR

Identification of polymers, plastics, contaminants,

organic films, fibers, and liquids

Infrared absorption Molecular chemical

identification

Molecular groups 0.1 - 100 ppm - No 15 microns

GC/MS GCMS

Identification and quantification of trace organic compounds

Molecular/characteristic fragment ions

Molecular ions -

400 ng (full scan)

10 ng (out gassing)

- - -

HFS Hydrogen in thin films (Quantitative)

Forward scattered hydrogen atoms H, D - 0.01 at% 50 nm No 2 mm x 10

mm

Raman Identification of organics and inorganics Raman scattering

Molecular chemical

identification

Molecular groups -

as low as 0.1 wt%

Confocal mode 1-2 µm Yes 1 µm

RBS Quantitative thin film composition and thickness Backscattered He atoms Li - U -

~5 at% (Z<20)~0.1 at%

(20<Z<70) ~0.005 at%

(Z>70)

2 - 20 nm Yes 2 mm

Table 2 Analytical Technique Summary (continued)

Analytical Technique Typical Applications Signal Detected Elements

Detected Organic

Information Detection

Limits Depth

Resolution Imaging/ Mapping

Lateral Resolution

(Probe Size)

SEM/ EDS Imaging and elemental microanalysis

Secondary and backscattered electrons and

X-rays B - U - 0.1 - 1 at% 1 - 5 micron

(EDS) Yes

4.5 nm (SEM)

1 micron (EDS)

FE SEM High resolution imaging of

polished precision cross sections

Secondary and backscattered electrons - - - - Yes 1.5 nm

SIMS Dopant and impurity depth

profiling, surface, and microanalysis

Secondary ions H – U - 1e12-1e16

at/cc (ppb-ppm)

5 30 nm Yes

1 micron (Imaging), 30 micron (profiling)

TEM High resolution imaging of thinned cross sections and

planar samples Transmitted electrons - - - - Yes 0.1 nm

TOF SIMS Surface microanalysis of polymers, plastics, and

organics

Secondary ions, atoms, molecules H – U Molecular ions

to mass 10,000 <1 ppma,

1e8 at/cm2 1 monolayer Yes 0.10 micron

TXRF Metallic contamination on semiconductor wafers Fluorescent X-rays S – U - 1e9-1e12

at/cm2 - Yes 10 mm

XPS/ESCA Surface analysis of organic and inorganic molecules Photoelectrons Li – U Chemical

bonding 0.01 - 1 at% 1 - 10 nm Yes 10 micron - 2 mm

XRD Crystal phase

identification, orientation, and crystallite size

Diffracted X-rays - - - - No 20 µm

XRF Thin film, thickness composition X-rays Na – U 10 ppm - No 100 µm

Determining the Appropriate Technique

There are numerous factors which can affect the decision of choosing the most appropriate technique. In many cases more than one technique may be applied, and trade-offs, such as improved spatial resolution versus lower concentration sensitivity, are involved. Several important factors are discussed in this section.

Required Spatial Resolution As shown in Figure 2, as the feature of interest shrinks in size, the number of available techniques for analysis drops significantly. In general, the only techniques that can normally resolve features that are below 1µm in size are electron-based techniques such as EDS, AES, SEM, and TEM, and the scanning probe (SPM) based techniques such as AFM. The disadvantage of these techniques is that only images (SEM, TEM. AFM), or elemental information (AES, SEM/EDS, TEM/EDS, TEM-EELS) are provided. Also, as higher spatial resolution is required, the time and cost associated with an analysis increases significantly. For example, SEM/EDS is typically the least expensive analysis, is usually available in-house, and can easily identify 1µm features such as particles, layers, or defects. For features below 1,000Å in size, high-resolution FE-AES is usually required, and for features that are less than 100Å, TEM/EDS or TEM-EELS is required. FE-AES analysis might cost ~2X that of SEM/EDS, while TEM based analyses can cost 5X-10X that of SEM/EDS due to the extensive sample preparation required to generate an electron-transparent sample. In contrast, AFM can provide high-resolution surface topography maps with a lateral resolution down to 100Å, and a vertical resolution down to 1Å, however, elemental identification cannot be provided as with the electron-beam based techniques.

Detection Sensitivity As shown in Figure 2, as the requirement for detection sensitivity drops below ~0.1% concentration the number of available techniques for analysis also decreases. In particular, Dynamic SIMS is the only technique available for measuring dopant level concentration profiles below the surface of materials. TOF-SIMS, TXRF, and VPD based techniques can provide high-sensitivity measurements for surface contaminants, but are not used for depth profiling. VPD involves chemically removing the native oxide on the wafer surface, so it is not possible to sample below the oxide, and the

physics of X-ray fluorescence effectively limit the useful sampling depth of TXRF to the upper ~100Å of a wafer. TOF-SIMS can be used for shallow depth profiling, however, because of lengthy analysis times profiles are typically limited to the upper ~100Å of a sample.

Sampling Depth Each technique has a different sampling depth, which sets the absolute depth resolution for that technique. For example, depths of 0.5-5µm are analyzed during XRF, EDS, and FTIR analysis, making detection of thin residues difficult with these techniques. AES, XPS/ESCA, and TXRF have a sampling depth of 30-100Å, making these techniques much more sensitive to surface contamination, while TOF-SIMS has the shallowest sampling depth, measuring the top 1-3 atomic layers of material, approximately 10Å, on a sample. Dynamic SIMS has a variable sampling depth depending upon the analysis conditions selected. Sampling depths may be as shallow as 10Å or as deep as 300Å. At the other extreme, Raman and FTIR typically require samples that are at least 0.2µm thick in order to obtain useful signal from the material. In some cases when sufficient material is not available, an organic solvent can be used to collect contaminant material from a large surface area of a sample and increase the analysis sensitivity.

Surface Versus In-Depth Analysis Analysis of buried layers and subsurface dopants, impurities, or defects requires a technique that can access the depth of interest. SIMS, AES, and XPS/ESCA perform quantitative depth profiling by sputtering the sample surface during analysis to expose sequentially greater depths in the sample. This allows virtually any depth to be investigated, although profiling of very thick (>10µm) structures can become very time consuming and cost prohibitive. Depth profiles can be obtained with TOF-SIMS, however, they are usually constrained to the upper ~100Å of a sample and are typically not quantified. RBS uses analysis of the energy of particles scattered from different depths in a sample to provide a non-destructive quantitative depth profile of the top ~1µm of samples. SEM, TEM, or AES cross section analyses use sample polishing or FIB (Focused Ion Beam) cuts to expose a layered structure for imaging and/or elemental analysis. AFM, FTIR, Raman, XRF, and TXRF do not have the capability of removing any material during analysis, and thus cannot provide depth profiles. However, Raman analysis can be performed in a confocal mode, allowing different analysis depths to be selected with a resolution of 1-5µm.

Conductive Versus Insulating Samples

While the electron-based techniques listed above do provide the best spatial resolution available, they are delivering a high-density current of electrons to the sample in a very small area. Charging can occur in insulating materials causing a loss of the analytical signal. For some techniques, this concern is not significant or has become a routine matter. In the case of SEM/EDS analysis, the sample can be coated with a conductive metal or carbon layer to eliminate charging. Metal layers tend to provide higher secondary electron yields and are better conductors than carbon layers. Carbon deposition involves a heated filament which can damage thermally sensitive samples, but carbon layers do offer advantages for EDS analysis since they allow better transmission of X-rays out of the sample. Also, metal coatings can generate several interfering X-ray peaks, while carbon layers do not. In TEM analysis the sample is usually sufficiently thin that only a small amount of charge is actually deposited in the sample. For SEM/EDS the sampling depth is on the order of 1µm, so any thin conductive layer that might be added typically adds only a small signal to the EDS spectrum. On the other hand, the AES analysis sampling depth is ~30Å, so deposition of a thin metal layer will mask the signal from the surface of interest, and is normally avoided. In extreme cases, such as analysis of 500Å particles on quartz photo masks, the only option may be to deposit a thin layer of metal on the mask, then sputter through the coating in a small area to expose a particle. In this manner a conductive path from the coating to ground is maintained. Analysis of bulk insulators such as quartz or plastics, or thick (>5,000Å) layers of materials like SiN, SiO2, polymide, or photoresist can make SEM/EDS and AES analyses difficult or impossible. While the presence of insulating materials does not automatically preclude the use of any technique, it is important to communicate their presence to the scientist performing the analysis, since this may effect sample preparation, or how the analysis is performed, if at all.

Elemental Versus Chemical Analysis There are numerous potential sources for organic contamination because organic materials and precursors are now an integral part of semiconductor devices, fabrication processes, and equipment. In addition, organic contamination has become an increasing problem with smaller device dimensions. With thousands of different organic materials present

in today’s environment, the identification of organic contaminants can be difficult. Unfortunately, the techniques that afford the highest spatial resolution generally only provide elemental information. As such, they may be able to confirm the presence of organic contamination, but not be able to identify the specific contaminant. Techniques such as TOF-SIMS, Raman, FTIR, and GC/MS can provide valuable information on the identity of organic materials. Unfortunately, none of the these techniques has sub-micron resolution, which can be frustrating when these sub-micron particles can be imaged by AES or SEM/EDS, but cannot be individually resolved by an organic technique. In some cases, identification of the contamination can still be obtained because techniques such as TOF-SIMS have very high sensitivity. This means that the signal from particles can still be detected, although they cannot be imaged. In this case, it is important to identify the location of high levels of particles or contamination along with comparable clean areas, or other clean samples. As long as a sufficient amount of material is present, comparison of the spectra between clean and contaminated areas can still provide the identification of the contaminant species, even though individual particles may not be resolved. XPS/ESCA analysis also offers the unique capability of being able to determine the chemical bonding state of assorted materials. This can be especially useful for interconnect problems where oxidation or corrosion of metal layers, such as bond pads, is an issue.

Quantitative Versus Qualitative Results With the exception of RBS, all of the analytical techniques discussed here require calibrated reference standards in order to provide high accuracy, quantitative results. Some techniques, such as AES and XPS/ESCA can provide semi-quantitative results by using recorded sensitivity factors. Other techniques, such as Dynamic SIMS and TOF-SIMS, are very matrix sensitive and require specific reference standards for the material being analyzed. Fortunately, large libraries of reference samples have been generated for many semiconductor materials, allowing quantitative results to be generated in many cases. All techniques can typically provide good relative comparisons between similar matrix samples, which may be all that is needed for comparison of good-to-bad samples. In general, quantitative results are preferred, so that results from different measurements performed at different times can be directly compared. For example, a 1x1019 atoms/cm3 concentration of B in a

Si matrix might generate signals varying between 1,000 counts to 100,000 counts in a SIMS profile depending upon the instrument setup used. Therefore, comparison of results obtained under different conditions will not be valid unless the results are calibrated using a reference standard. If results are calibrated and presented in concentration versus depth format, then results obtained months or even years apart can be compared, as long as correct analytical procedures for quantification were followed.

Destructive Versus Non-destructive Techniques Techniques such as AES, XPS/ESCA, SEM, FTIR, Raman, RBS, and XRF can usually analyze a sample surface without causing significant modification. Other techniques, such as Dynamic-SIMS and AES or XPS depth profiling consume sample material during analysis, and thus are destructive. The destructive properties of a technique need to be considered if one is trying to analyze a unique defect or failure. Another consideration is sample size requirements of the technique. Some instruments can analyze full 200mm or even 300mm wafers, while others such as TEM require very small samples thus requiring significant modification of the sample during preparation. Also, if more than one analysis is to be performed on a single sample, then consideration should be given to the order in which analyses are performed. For example, when a SEM analysis is performed, a thin carbon layer is often deposited from the ambient vacuum onto the sample surface during the analysis. Also, the high-density electron beam used for SEM analysis can modify fragile organic materials and oxides. This can result in a loss of valuable organic information such that subsequent TOF-SIMS analysis may not be able provide useful results on organic contamination if a defect has already been analyzed by SEM. Similarly, if a sample has been metal coated in order to make it conductive for SEM/EDS analysis, it is futile to perform TOF-SIMS analysis for organic contamination, because the metal coating would cover any organic layer that might have been present.

Good Versus Bad Analysis In many cases, the best method for determining a failure mechanism is to compare bad versus good samples. One commonly encountered problem is that good samples are not available for comparison, because all of the good ones are already in products that have been shipped out of the fab. If a process is working well, then it may be prudent to save typical samples from different steps along the process flow. For example, if the profile shape of a particular ion

implant is crucial in controlling a device threshold voltage, then having a blanket wafer implanted with only this implant when the process is in control would allow direct comparison to a bad wafer when the process is out of control. In many cases engineers are surprised to find that the structure of “good” samples is not what they expected. If a good sample is not available for comparison, then normal features in a bad sample may be incorrectly interpreted as the source of the problem.

Analysis of Specific Elements Certain techniques cannot detect all elements. EDS, AES, and XPS/ESCA cannot detect H and He, and TXRF cannot practically detect elements lighter than S. In other cases, spectral interferences can limit the detection sensitivity or the ability to quantify certain elements in a specific sample matrix. For example, in AES analysis one of the peaks for Ti coincides with the only peak for N, thus limiting the absolute accuracy for the analysis of TiN films by AES.

Survey Versus Specific Element Analysis Techniques such as AES, EDS, XPS/ESCA, TOF-SIMS, and XRF/TXRF are good “survey” techniques. That is, they can detect a wide number of elements that could be present on your sample. Depth profiling techniques such as Dynamic SIMS, AES, or XPS/ESCA which use sputtering for depth profiling, require that the elements to be analyzed are specified before the profile is started, and typically are limited to analyzing no more than 8 elements in a single profile. This can present a problem when searching for unknown contaminants that are buried in a sample such as impurities in a gate oxide layer. There is no easy way to scan across a wide range of elements when trying to detect dopant level impurities at a buried interface in a sample. The most reasonable approach is to prioritize potential contaminants by the likelihood of their presence, and then obtain profiles while monitoring those elements. Electrical results or other information is often helpful in narrowing the list of potential contaminants, however, it may still be necessary to acquire several profiles in order to pinpoint the impurity causing the problem.

Cost Considerations In-house techniques are usually considered to be “free”, so the cost of outsourcing is an important issue. Careful consideration of the goals and subsequent experiment design in consultation with the scientist performing the experiment can yield

significant cost savings. For example, TEM analysis has historically been needed to accurately measure the thickness of layers that are less than a few hundred Ångstroms in thickness. With the advent of In-Lens Field Emission SEM systems, the range of SEM has been expanded. In many cases, now SEM can measure films as thin as 50-100Å. Unlike the TEM, SEM analysis does not require thinning a sample to electron transparency, and so a significant cost saving can be realized. On the other hand, there is not always an inexpensive analysis option available. For example, the problem being investigated may be complex, such as multiple potential contaminant sources, or require the very best spatial resolution, such as thin oxide formation or alloying between interconnect layers. The key to cost control is not to ask for more than what is required to answer the relevant questions.

Availability of Techniques and Instruments

Numerous commercial sources are available for techniques such as SEM, TEM, FTIR, Raman, AES, XPS/ESCA, and XRF. The sources for Dynamic SIMS and TOF-SIMS, TXRF, and RBS are less numerous, yet there are still several labs around the world that offer these techniques. TEM-EELS and GDMS are only commercially available at one or two labs. While techniques such as TEM, AES and XPS are widely available, there are significant differences in instrumentation capabilities such as absolute spatial resolution, or full wafer capability. Also, faster turnaround may be available from alternate sources for techniques that are widely available. If there is only one source for a particular measurement, then the turnaround time for that analysis may vary significantly depending upon the sample backlog at any given time.

Comparability with In-house Techniques Analysis protocols and conditions can vary widely depending upon the goals of analysis. Just because an analysis was performed using the same technique, or even the same model of instrument does not guarantee that the results will be comparable with those obtained in-house. For example, during SIMS analysis of mobile ions in SiO2, Na will migrate to the SiO2/Si interface unless the sample is properly charge compensated during the analysis. In one case, in-house analyses were performed using protocols that were developed for an instrument that could not perform proper charge compensation, so the profiles always showed Na as being present only at the interface. Tabulated “Interfacial Na” results actually reflected the total

amount of Na present in the SiO2, and these results were used for process control. The work was outsourced with the goal of the analysis stated only as “Measure Na at SiO2/Si interface”, so the analysis was performed using proper charge neutralization and an accurate profile was obtained indicating most of the Na was actually present within the SiO2 layer and not at the interface. Confusion ensued because the tabulated interfacial Na levels obtained from a control sample were orders of magnitude lower than an in-house, uncompensated analysis from the same sample. Also, because of quantification artifacts associated with the uncompensated measurement it was not even possible to compare the total integrated Na concentrations from the charge compensated measurement. Since the Na analysis was being used for process control it was essential that the new results be directly comparable to the historical results. This meant that the entire measurement had to be repeated using exactly the same protocol as the historical in-house measurements. If samples are being outsourced because an in-house lab is overloaded or equipment is down, then one would naturally tend to use the same technique, and possibly even the same instrumentation or protocols, as the in-house lab in order to allow the outsourced results to be compared to the in-house results. It is therefore crucial to communicate to the scientist performing the work what conditions were used during the previous analyses and that you want to compare these results to those previously obtained, e.g., for process monitoring. If samples are being outsourced because the in-house expert is sick or on vacation, and results are needed before they return, then the external scientist will have to do the best they can, and will need as much information as possible. If samples are being outsourced because the in-house staff are overwhelmed, then careful consideration should be given to which samples are sent out. If your in-house scientist has developed specialized proprietary protocols, or is the only person in the world who knows how to do a certain analysis, then those samples should remain in-house if at all possible, and other samples requiring more routine analysis should be outsourced. One should also consider including a sample that was previously analyzed in-house in order to confirm the comparability of the external results to in-house results. Many techniques require the use of calibration samples in order to provide quantitative results, and the accuracy of calibration samples can vary significantly. Commercial laboratories tend to go to greater lengths to insure the accuracy of their calibration standards since their results are often used to qualify production tools with specific

requirements. Internal standards tend to be used to insure reproducibility of results used for process control and therefore less significance is placed upon the absolute accuracy of these standards.

Ask an Expert A wealth of advice and information is available from experts within the commercial laboratories. Typically, senior scientists or the technical manager can provide good feedback on whether a given technique can answer a specific question. Some labs now have technical customer service representatives whose job is to work with customers in designing experiments and determining which technique is most suited to their need. These scientists are selected because they have a breadth of experience in working with numerous techniques, and can provide a more objective opinion on the selection of an appropriate technique. Good commercial laboratories want repeat customers and understand that if an inappropriate technique is recommend, the odds of a return customer are reduced. This could mean recommending a different technique, another lab, or even turning away work completely if it is clear that none of the available techniques are appropriate. While the issues above may seem numerous and complicated, in general, an appropriate analysis technique can be determined quite quickly with input from an experienced scientist. Summarized below is a listing of the techniques used to analyze specific problems. Surface versus Buried Features Surface Features: Particles – <0.1 µm: FE-AES – <1 µm: SEM-EDS, FE-AES – <10 µm: SEM-EDS, FE-AES, TOF-SIMS, Raman – >10 µm: SEM-EDS, FE-AES, TOF-SIMS, Raman µ-FTIR, µ-XPS Residues – Inorganic: SEM-EDS, FE-AES, XPS, TOF-SIMS – Organic: µ-FTIR, XPS, TOF-SIMS, TXRF Stains, discolorations or hazes – SPM, SEM (physical characterization) – XPS, AES, TOF-SIMS, FTIR, SEM-EDS, TXRF Low-concentration metallic and dopant surface contamination –TOF-SIMS, TXRF, Dynamic-SIMS, VPD-ICPMS/GFAAS Surface Roughness, Morphology, and Feature Size –AFM, SEM

Buried Features: Buried Defects, Particles, and Layers (in conjunction with polished or FIB cross sectioning, or Ar sputter depth profiling for FE-AES): – <0.01 µm: TEM-EDS, TEM-EELS – <0.1 µm: FE-AES – <1 µm: SEM-EDS, FE-AES – <10 µm: SEM-EDS, FE-AES Dopant and Trace Level Contaminant Profiling – Dynamic-SIMS Thin Film Depth Profiling – <10 µm: FE-AES – <100 µm: AES, XPS, Dynamic-SIMS – Blanket films: RBS, AES, XPS, Dynamic-SIMS Survey versus Specific Element Techniques “Survey” Techniques: AES, EDS, XPS/ESCA, TXRF, TOF-SIMS, FTIR, Raman, GC/MS, XRF, GDMS Specific Element Techniques: Dynamic SIMS, AES depth profiling, XPS/ESCA depth profiling, RBS/HFS

Information to Provide When Submitting Samples

A surprising number of samples arrive at laboratories without any kind of identifying feature on them other than being marked “RUSH!” In their haste to get samples out the door for rush analysis, people often forget to even put their name and phone number on the box or in with the samples. At larger labs up to a hundred sets of samples can arrive in a single day, with several of them being for rush turnaround. The guidelines shown below can improve your chances of getting faster turnaround and a quality analysis.

Contact Information The necessary information should include your name, company, address, office phone, mobile phone, pager number, and email address, along with information for an alternate contact. In many cases a scientist will observe something unexpected during an analysis, such as extra or missing layers. If the appropriate individual can be contacted while the analysis is being performed, the experiment may be modified or terminated as appropriate. In the worst case, if absolutely no identification is included with samples, delays of several days can occur while the identity of the client is determined and they are contacted.

Purpose of the analysis Try to be as explicit as possible. Include photos, diagrams, maps, and previously obtained results. If the analysis is to be performed in a specific microscopic location, then photos or maps with gradually increasing magnification or instructions referencing easily located landmarks are useful.

Priority of Samples Most analyses are time consuming, and only so many can be done in one day or one batch. In cases involving large numbers of samples it may be possible to provide results from the top priority samples before the entire set of the analysis is completed.

Expected Concentrations Analysis conditions are often modified to provide optimum sensitivity, and high concentrations of elements may actually saturate detection systems and lead to inaccurate results. A good scientist will repeat an analysis that had initially resulted with a saturated signal, however, it is preferred the analysis be initiated with the proper sensitivity from the start.

Required Sensitivity, Depth of Analysis If you are concerned about an impurity at very low levels, or need to analyze to a certain depth in a sample, state so explicitly. For example, if you are only interested in the concentration profile of an element in the upper 0.1µm of a 1µm film it may be possible to modify the experiment so as to provide a profile of only the top 0.1µm with enhanced depth resolution in this depth range.

Previous Analyses If previous analyses have been performed and you want to compare results from the new samples, then a reference to the earlier work should be included so that similar analytical conditions can be used.

Avoid the Use of Jargon and Acronyms Everyone at XYZ Corporation may know what a BFD987 structure or what an HFC-9847 test procedure is, but the scientist doing your analysis probably does not and will not be able to proceed with the analysis until they determine what is meant. In many cases a simple sketch, a photo, or layered diagram are sufficient to clarify the situation.

Tricks of the Trade Improving the quality of the analysis and reducing response time while controlling cost are important considerations when submitting samples for analysis. Below are some ways to improve chances of success, control cost and cut turnaround time:

“Black box” versus Expert Analysis The scientist who is performing the analysis has skills and knowledge that can be a valuable asset but only if allowed to take advantage of them. Clients often want to provide the absolute minimum amount of information because of concerns that divulging specific information about each sample will bias the results. For example, films deposited on Samples 1 through 10 might be expected to have increasing film thickness with increasing sample number. If the scientist observes that Sample 6 is thicker than Sample 7, and they know that this does not follow the expected trend, then they may repeat the analysis of these two samples and double check the identification of those samples in order to confirm the results. If the scientist is treated like a “black box”, and told only to determine the thickness of the film on each sample, then they have no way of knowing that the results for Samples 6 and 7 don’t make sense and would not double check those samples. Skeptics might be concerned that the scientist performing the above experiment might change the sample identifications to match the expected results. However, in 20 years of experience this writer has never witnessed a single case of results being intentionally manipulated in order to match a customer’s expectation. Also, if the scientist is aware of the goal behind the experiment, he may be able to add insight to the results. For example, in the case described above, the scientist might notice other differences in Samples 6 such as high contamination levels or film roughness, that might help explain why it did not fall in the expected order. On the other hand, the analyst should not be expected to provide a complete solution to your problem. Their job is to provide scientific data from a specific experiment and lend insight into the validity and accuracy of those measurements. They might be able to relate the data to process or contamination problems, but the final solution to the problem is ultimately the customer’s responsibility.

End Point Analysis Analysis of sets of samples generated when a tool has been operated over a range of conditions is often requested in order to generate a calibration or

performance curve based upon the operating conditions. In some cases, virtually no difference is detected across the entire range of samples, with a significant cost incurred to determine that the samples were all the same. A beneficial strategy would be to select a subset of samples from the extremes of the experiment and specify that these samples be analyzed first. If there is no significant difference between the extremes, the remaining samples need not be analyzed. This is a case where using the analytical skills of the scientist who is performing the experiment can save a significant amount of time and money. If submitted as a “black box” experiment, the scientist has no option but to analyze (and charge) for all the samples.

Iterative Analysis In some cases it may not be clear if a particular technique can provide useful results, but still an entire batch of samples may be sent to the lab. Only request analysis on one or two of them first. Once the results from those samples have been received and it has been determined that the analysis is providing useful results, then the analysis of the rest of the samples can be authorized. Using this approach will take longer to deliver the results from all of the samples, but can prevent spending of a lot of money for results that don’t address the problem. Conversely, if time is of the essence and the problem has to be solved immediately, one may choose to perform several analyses simultaneously in order to minimize the wait for results. Only one of the techniques may provide useful results, but the solution to the problem will be found the quickest.

Rush Analysis Results are often needed urgently, and most commercial labs offer expedited turnaround for a surcharge. Unfortunately, the analysis can’t start until the samples arrive at the laboratory. Even if a surcharge is not being paid, it is likely that results are desired back as quickly as possible. Making appropriate contacts in your company’s shipping department or shipping the samples yourself can shave days off of turnaround times for analysis. In many cases a client will intend to ship samples by overnight courier to arrive the next day, but the samples show up 2, 3, or even 4 days later because the shipping department was backlogged or because the samples were shipped by U.S. mail or ground delivery instead of overnight air delivery. This might save $20 on the shipping cost, but add hundreds or thousands of dollars to the analysis costs because a rush surcharge had to be applied in order to provide the results before a specific deadline. Finally, as a

common courtesy, if an urgent “top priority” analysis has been arranged and then samples were not sent, let the lab know that the request has been canceled. It is very frustrating and costly for a scientist to have an instrument set up and idly waiting for samples to arrive at 9AM only to find out at noon that they are not coming at all.

Summary A wide range of outsourcing options is now available for surface and microanalysis. Judicious use of these resources can help solve failure analysis and other materials and processing problems in a timely and cost-effect manner. The requestor has the final decision as to which technique is used and how it is applied to answer any analytical questions, and so the key to successful outsourcing is thorough and succinct communication to and from the outsourcing laboratory.