What are quality agreements, why do we need them, who should be involved with writing them, and what should they contain?

In May, the Food and Drug Administration (FDA) published new draft guidance for industry entitled "Contract Manufacturing Arrangements for Drugs: Quality Agreements" (1). I just love the way the title rolls off the tongue, don't you? You may wonder what this has to do with spectroscopy or indeed an analytical laboratory, which is a fair question. The answer comes on page one of the document where it defines the term "manufacturing" to include processing, packing, holding, labeling, testing, and operations of the "Quality Unit." Ah, has the light bulb been turned on yet? Testing and operations of the quality unit — could these terms mean that a laboratory is involved? Yes, indeed. 

Paddling across to the other side of the Atlantic Ocean, the European Union (EU) issued a new version of Chapter 7 of the good manufacturing practices (GMP) regulations that became effective on January 31, 2013 (2). The document was updated because of the need for revised guidance on the outsourcing of GMP regulated activities in light of the International Conference on Harmonization (ICH) Q10 on pharmaceutical quality systems (3). The chapter title has been changed from "Contract Manufacture and Analysis" to "Outsourced Activities" to give a wider scope to the regulation, especially given the globalization of the pharmaceutical industry these days. You may also remember from an earlier "Focus on Quality" column (4) dealing with EU GMP Annex 11 on computerized systems (5) that agreements were needed with suppliers, consultants, and contractors for services. These agreements needed the scope of the service to be clearly stated and that the responsibilities of all parties be defined. At the end of clause 3.1 it was also stated that IT departments are analogous (5) — oh dear!  

Also, ICH Q7, which is the application of GMP to active pharmaceutical ingredients (APIs), has section 16 entitled "Contract Manufacturing (Including Laboratories)." This document was issued as "Guidance for Industry" by the FDA (6) and is Part 2 of EU GMP (7). Section 16 states that "There should be a written and approved contract or formal agreement between a company and its contractors that defines in detail the GMP responsibilities, including the quality measures, of each party." As noted in the title of the chapter, the scope includes any contract laboratories, which means that there needs to be a contract or formal agreement for services performed for the API manufacturer or any third-party laboratory performing analyses on their behalf. 

Therefore, what we discuss in this gripping installment of "Focus on Quality" is quality agreements, why they are important for laboratories, and what they should contain for contract analysis. The principles covered here should also be applicable to laboratories accredited to International Organization for Standardization (ISO) 17025 and also to third-party providers of services to regulated and quality laboratories.   

 is a comprehensive written agreement that defines and establishes the obligations and responsibilities of the quality units of each of the parties involved in the contract manufacturing of drugs subject to GMP (1). The FDA makes the point that a quality agreement is not a business agreement covering the commercial terms and conditions of a contract, but is prepared by quality personnel and focuses only on the quality of the laboratory service being provided. 

Note that a quality agreement does not absolve the contractor (typically a pharmaceutical company) from the responsibility and accountability of the work carried out. A contract laboratory should be viewed as an extension of the company's own facilities. 

Put at its simplest, the purpose of a quality agreement is to manage the expectations of the two parties involved from the perspective of the quality of the work and the compliance with applicable regulations. Under the FDA there is no explicit regulation in 21 

 211 (8), but it is a regulatory expectation as discussed in the guidance (1).  

				In contrast, in Europe the requirement is not guidance but the law that defines what parties have to do when outsourcing activities as (2):

Any activity covered by the GMP Guide that is outsourced should be appropriately defined, agreed and controlled in order to avoid misunderstandings which could result in a product or operation of unsatisfactory quality. There must be a written Contract between the Contract Giver and the Contract Acceptor which clearly establishes the duties of each party. The Quality Management System of the Contract Giver must clearly state the way that the Qualified Person certifying each batch of product for release exercises his full responsibility. 

Who Is Involved in a Quality Agreement? (Part I)

Typically, there will just be two parties to a quality agreement; these are defined as the contract giver (that is, the pharmaceutical company or sponsor) and the contract acceptor (contract laboratory) according to EU GMP (2). Alternatively, the FDA refers to "The Owner" and the "Contracted Facility" (1). Regardless of the names used, there are typically two parties involved in making a quality agreement unless, of course, you happen to be the Marx Brothers (the party of the first part, the party of the second part … the party of the 10th part, and so forth) (9). 

If the contact acceptor is going to subcontract part of their work to a third party, then this must be included in the agreement — with the option of veto by the sponsor and the right of audit by the owner or the contracted facility. 

Figure 1 shows condensed requirements of EU GMP Chapter 7 regarding the contract giver or owner and the contract acceptor or contracted facility to carry out the work. Responsibilities of the owner are outlined on the left-hand side of the figure and those of the contracted facility are outlined on the right-hand side of the diagram. The content of the contract or quality agreement is presented in the lower portion of the figure. These points will be discussed as we go through the remainder of this column.  

 Figure 1: Summary of EU GMP Chapter 7 requirements for outsourced activities.			

According to the FDA (1), a quality agreement has the following sections as a minimum: 					

Terms (including effective date and termination clause)

Responsibilities, including communication mechanisms and contacts

Although quality agreements can cover the whole range of pharmaceutical development and manufacturing, for the purposes of this column we will focus on the laboratory. In this discussion, the term laboratory will apply to a laboratory undertaking regulated analysis either during the research and development (R&D) or manufacturing phases of a drug product's life, including instances where the whole manufacturing and testing is outsourced to a contract manufacturing organization (CMO), the whole analysis is outsourced to a contract research organization (CRO) or testing laboratory, or just specialized assays are outsourced by a company. In fact, there are specific laboratory requirements that are discussed in the FDA guidance (1). 

The scope of a quality agreement can cover one or more of the following compliance activities: 					

Qualification, calibration, and maintenance of analytical instruments. It may be a requirement of the agreement n that only specified instruments are to be used for the owner's analysis.

Validation of laboratory computerized systems. 

Validation or verification of analytical procedures under actual conditions of use. This will probably involve technology transfer, so the agreement will need to specify how this will be handled.

Specifications to be used to pass or fail individual test results.

Supply, handling, storage, preparation, and use of reference standards. For generic drugs, a United States Pharmacopeia (USP) or European Pharmacopoeia (EP) reference standard may be used, but for some ethical or R&D compounds the owner may supply reference substances for use by a contract laboratory.

Handling, storage, preparation, and use of chemicals, reagents, and solutions.

Receipt, analysis, and reporting of samples. These samples can be raw materials, in-process samples, finished goods, stability samples, and so on.

Collection and management of laboratory records (for example, complete data) including electronic records (10) resulting from all of the above activities.

Deviation management (for example, out-of-specification investigations) and corrective and preventative action plans (CAPA).

Change control: specifying what can be changed by the laboratory and what changes need prior approval by the owner. 

All activities under the scope of the quality agreement must be conducted to the applicable GMP regulations.  

More Detail on Sections in a Quality Agreement 

Now, I want to go into more detail about some selected sections of a quality agreement. I would like to emphasize that this is my selection and for a full picture readers should refer to the source guidance (1) or regulation (2).  

In the majority of instances the owner or contract giver will supply their own analytical procedures and specifications for the contract laboratory to use. These will be specified in the quality agreement. Part of the quality agreement should also define who has the responsibility for updating any documents. If the owner updates any document within the scope of the agreement, then they have a duty to pass on the latest version and even offer training to the contract facility. To ensure consistency, the owner could require the contract laboratory to follow the owner's procedure for out-of-specification results. 

The responsibilities of the two groups signing the agreement will depend on the amount of work devolved to the contract laboratory and the degree of autonomy that will be allowed. For example, if a laboratory is used for carrying out one or two specialized tests rather than complete release testing, then the majority of the work will remain with the owner's quality unit. The latter will handle the release process and the majority of work. In contrast, when the whole package of work is devolved to a CMO, then the responsibilities will be greatly enlarged for the contract laboratory and communication of results, especially exceptions, will be of greater importance.  

In any case, the communication process between the laboratory and the owner needs to be defined for routine, escalation, and emergency cases. How should the communication be achieved? The choices could be phone, fax, scanned documents, e-mail, on-line access to laboratory systems, or all of the above. However, it is no good having the communication processes defined as they will be operated by people, therefore analysts, quality assurance (QA) staff, and their deputies involved on both sides have to be nominated and understand their roles and responsibilities.    

Communication also needs to define what happens in case of disputes that can be raised from either party. The agreement should document the mechanism for resolving disputes and, in the last resort, which country's legal system will be the beneficiary of the financial largesse of both organizations as they fight it out in the courts. Unsurprisingly, I'm in the camp with the great spectroscopist "Dick the Butcher" who said "The first thing we do, let's kill all the lawyers" (11). 

Just in case you missed it above, the contract laboratory must generate complete data as required by GMP for all work it undertakes. This will include both paper and electronic records that I discussed in an earlier "Focus on Quality" column (10). The FDA guidance makes it clear that the quality agreement must document how the requirement for "immediate retrieval" will be met by either the owner or the contract laboratory (1). Both types of records must be protected and managed over the record retention period. 

Articles

Focus on Quality

In pharmaceutical contract manufacturing, the work of analytical scientists is covered by a quality agreement, which is prepared by personnel in the quality control or quality assurance department and focuses on the laboratory analyses provided. But what exactly are quality agreements, why do we need them, who should be involved in writing them, and what should they contain? Here, we answer those questions.

Why Do We Need Quality Agreements?

Raman spectroscopy has developed into a key analytical tool for the pharmaceutical industry. Both portable and benchtop Raman systems are now widely used for applications in the pharmaceutical industry. Despite a challenging economy, demand from the pharmaceutical industry is holding its own and should return to typical growth rates by next year.  

 Laboratory and portable Raman - pharmaceutical vendor share, 2012.			

  Within the pharmaceutical industry, one of the biggest applications now is the identification of incoming raw materials and finished product inspection, for which there are numerous new portable and handheld models available for on-dock or production area analysis. The other major application is the analysis of polymorphs and salt forms in both quality control (QC) and product development, variations of which can lead to dramatically different performance of a pharmaceutical product. 

    Worldwide demand for laboratory and portable Raman spectroscopy instrumentation from the pharmaceutical industry was about $36 million in 2012. Global economic conditions and major cuts in government spending will clearly make 2013 a challenging year for the overall Raman market. However, the numerous new product introductions and market entrants over the last two years will help keep demand from the pharmaceutical industry in positive territory in 2013, if only slightly, until much stronger growth returns, most likely in 2014. 

    The foregoing data were extracted from SDi's market analysis and perspectives report titled 

The Global Assessment Report, 12th Edition: The Laboratory Life Science and Analytical Instrument Industry, October 2012

Departments

Raman spectroscopy has developed into a key analytical tool for the pharmaceutical industry.

Market Profile: Raman Spectroscopy in the Pharmaceutical Industry

The time is ripe for further recognition of the breadth of analytical mass spectrometry. Other than awards, prizes, and fellowships, an alternative method of recognition in the community is to use an individual's name in a description of the contribution — an eponymous tribute. Eponymous terms enter into the literature independently of prize or award, and are often linked to a singular, specific, and timely achievement. Here, we highlight the Brubaker prefilter and Kendrick mass.

The 2013 annual meeting of the American Society for Mass Spectrometry (ASMS) concluded a few months ago. The continued growth of that conference, and the breadth of research and application in mass spectrometry (MS) evident there and described in other professional research conferences, reflects the vitality of modern MS. That growth seems to continue with few limits. The growth is catalyzed by a need for analysis in new application venues at better sensitivities, but growth is also supported by innovation in instrumentation and information processing. The 2013 ASMS conference celebrated 100 years of MS, and Michael Gross gave a presentation "The First Fifty Years of MS: Building a Foundation." It is still an unattainable ideal to reliably forecast the future simply by examining the past, and truly significant advances often seem to appear from nowhere. As is true in most scientific research, MS develops predominantly through incremental improvements. We can generally predict such improvements. More singular revolutionary advances, still recognized over the perspective of years, should be recognized through research awards such as the Nobel Prize. A third type of advancement lies between the two, and is neither simply iterative (which belongs to everyone in a sense) or transformative (which is recognized by a singular prize such as the Nobel). These important achievements are recognized within the community by awards of professional organizations, but also through high citation in the literature of specific publications as well as the eponymous citation. Eponymous citations, interestingly enough, sometimes do not include a traditional citation and reference to a publication. Often, the use of a name stands alone. For example, in recent columns in this series, we have described the use of Girard's reagents in derivatization, Penning ionization, and the Faraday cup detector. Current scientific publications that use that method or those devices may not cite all the way back to the original publication, and may not include any citation at all. Using the name alone seems to suffice.  

In some fields of science, such as organism biology, the naming rights for a newly discovered organism go to its discoverer, and the scientist's name can be part of the term eventually approved by scientific nomenclators. In astronomy, naming rights for celestial bodies may also link to the first discoverer, or organizations may provide recognition of past achievements through the use of names. Features on the Moon and Mars, for example, reflect significant past scientific achievements. The instruments still roving Mars are providing a large number of "naming opportunities," some of which seem to be tongue-in-cheek. In chemistry, organic chemists who develop a certain reaction often see their names associated with that reaction (1–3). Ion fragmentation reactions in MS can follow this tradition; the McLafferty rearrangement is a significant eponymous example. In instrumentation and engineering, devices and mechanical inventions are also often named after their inventors (4). Sometimes, even when the inventor shuns publicity (5), the device is so perfect for its application that it remains in use for decades. Thus, users encounter the Brannock device without ever knowing the name, perhaps, but the cognoscenti are familiar with its history.  

Mass spectrometers are complex instruments, encompassing technologies from vacuum science, ion optics, materials science, electronics, information and computer science, and more. Instrumental MS showcases a number of inventions named after the inventor. We describe some here in this column. These represent eponymous terms that are, like the Brannock device, so widely used or are such an integral part of the science that they are referred to by name without citation or explanation. Table I includes a few eponymous terms that should be familiar to mass spectrometrists. In some cases, as for the several eponymous terms containing the name of A.O.C. Nier, the pioneering contributions have been described in honorific publications (6). The list in Table I is not comprehensive and does not include underlying eponymous terms in basic science such as Taylor cone, Fourier transform, Coulomb's law, Franck-Condon transitions, or the like. Some eponymous terms are in limited use or they may appear in the literature only a few times; these occurrences are both hard to discover and difficult to document. From the perspective of 50 or 100 years in laying the foundation for MS, there may be a few things that we have always known that we did not actually know until somebody pointed it out. The goal in this column is not to revisit some obscure contribution, but rather to provide an appreciation of contributions that support modern MS.  

Table I: A sampling of other eponymous terms in mass  spectrometry			

The basic concept of the quadrupole mass filter and the quadrupole ion trap was first disclosed in 1953; it was 36 years later that the contribution was recognized by the award of the 1989 Nobel Prize in Physics. Contrast that period of 36 years with the shortened period of only a few years between the awarding of the Nobel prize and discovery of deuterium and the neutron, as was described in a recent "Mass Spectrometry Forum" column in July 2013. By 1989, successful quadrupole-based commercial instrumentation had been on the market for almost 20 years. The rapid rise of gas chromatography coupled with mass spectrometry (GC–MS), and later liquid chromatography with mass spectrometry (LC–MS) can be linked directly to quadrupole devices. Quadrupole mass filters and quadrupole ion traps represented a fundamental change — they were smaller, cheaper, and their performance measures dovetailed with the needs of the hyphenated coupling. 

The Nobel prize recognized the transformative accomplishment (7). Continuous design innovations over many years created a more capable and reliable instrument, and these are described in an overview by March (8), selected from among the many that are available. In concordance with the theme of personal developments in the advancement of mass spectrometry, Stafford's retrospective (9) discusses the development of the ion trap at one commercial vendor.  

Let's consider the four-rod classical quadrupole mass filter. A key concept of this mass analyzer is its operation through application of radio frequency (rf) and direct current (DC) voltages to create "regions" of ion stability and instability within its structure. The term region applies to the physical volume subject to the electronic fields and gradients; each physical volume in the device is a certain distance from each of the four rods that make up the quadrupole mass filter, and a certain distance from either the source or the detector end, and the electric fields can be controlled. In simple terms, ions that pass through the filter from the source to the detector remain within stable regions. Those ions that enter regions of instability do not; they are ejected from the filter entirely or collide with the rods themselves. The quadrupole rods may be manufactured with different dimensions and the rods may be of various lengths. The applied fields vary in frequency and amplitude. Each of these factors influences the performance metrics in predictable ways. Given an entering ion with known kinetic energy, and a defined known trajectory, the performance of the quadrupole mass filter is predictable.  

However, ions (plural) are an unruly lot. It is not one ion that needs to pass through the quadrupole, but hundreds of thousands of ions, created in an ion source with its own physical design and operational characteristics. Therefore, the population of ions exiting the source and entering the mass analyzer has a spread of ion kinetic energies and a range of ion trajectories, as well as a diversity of ion masses and charge states. The stable ions may represent just a small fraction of that overall population, which would limit the transmission of the filter and result in decreased sensitivity for the analysis. As in other mass analyzers used in mass spectrometry, we might use lenses, slits, or other sectors (such as an electric sector in a double focusing geometry instrument) to tailor the kinetic energies and trajectories, and pass a larger fraction of the ion population into the mass analyzer. 

Or, in quadrupoles, we might use a Brubaker prefilter (sometimes called a Brubaker lens) situated in front of the mass-analyzing quadrupole. Wilson M. Brubaker was granted patent 3,129,327 (and later a related patent 3,371,204) for this device; the first patent was filed in December 1961 and granted in April 1964 (see Figure 1 for the first page of this granted patent). Brubaker worked for Consolidated Electrodynamics Corporation in Pasadena, California in the early 1960s. He was active in fundamental studies of the motions of ions in strong electric fields and in the development of small quadrupoles used to study the composition of the Earth's upper atmosphere. He also worked for Analog Technology Corporation and was the chief scientist for earth sciences at Teledyne Corporation. It was at this last position that Brubaker developed a miniature mass spectrometer (based on a quadrupole) for breath analysis for astronauts, which garnered a newspaper story on October 31, 1969. Brubaker's photograph that accompanied the article shows him posing with the sampling port of the mass spectrometer, and the photo is now available on eBay!  

Figure 1: A figure from Brubaker’s original patent, showing three of the four rods in the prefilter. The ions move along the z direction from lower left to upper right, and into the main structure of the quadrupole mass filter.			

A Brubaker prefilter (10) (Figure 2) consists of a set of four short, cylindrical electrodes (sometimes called stubbies) mounted colinearly and in front of the main quadrupole electrodes. Without the prefilter, ions from the source may not enter the quadrupole because of fringing fields at the front end, which are created by the exposed ends of the rods themselves. Transmission decreases because of this front-end loss. The Brubaker prefilter is driven with the same AC voltage (that is, the radio frequency voltage) applied to the main rods, and acts to guide ions in to the main quadrupole. It is an "rf-only" quadrupole that was later used in triple-quadrupole MS-MS instruments. The dynamics of these devices have been widely studied (11,12) and they serve several functions in the triple quadrupole instrument, depending on the pressure within the device. With the use of a Brubaker prefilter in a single-quadrupole instrument, the transmission of ions is greatly increased, although the increase is known to be mass-dependent. These devices are almost universally designed into quadrupoles. It is interesting that Brubaker's invention of the prefilter was a consequence of his need to maintain ion transmission in very small quadrupoles, such as what might be carried aboard the probes into the upper atmosphere or in spacecrafts. Transmission losses because of fringing fields are especially severe in these small devices. Nowadays, quadrupole mass filters can be microfabricated, with the prefilter and filter assembly only 3.2 cm in length. Wright and colleagues describe the incorporation of the Brubaker prefilter in miniature quadrupole mass filters (13).   

Figure 2: A simplified schematic (looking along the y-axis) of a Brubaker prefilter and the main quadrupole structure.			

As Table I shows, devices used in MS are often linked to the names of their inventors or designers. Modern MS is not only about the instrumentation, but also about the immense amounts of data generated by those instruments. From the first compilations of mass spectra in a three-ring binder from the American Petroleum Institute to the Eight Peak Index of Mass Spectral data (designed for searching of the most intense peaks in the mass spectrum) to larger modern databases, scientists have worried how to archive, present, and search mass spectral data. Mass spectral data are always at least two-dimensional (2D) (the 

 of an ion and its abundance). Time (as in information recorded with elution time in GC–MS or LC–MS) adds another dimension. Both higher resolution data and MS-MS data add dimensionality. A recent "Mass Spectrometry Forum" column (14) discussed the data management issue in MS. Large data sets can be mined for both their explicit first-order content and other meaningful secondary measures can sometimes be extracted. Insightful means of data presentation help tame the data glut. For example, time-resolved ion intensity plots for multiple-reaction-monitoring data highlight the underlying pattern. Three-dimensional (3D) color-coded plots for MS-MS data provide a portrait of mixture complexity and reactivity. Statistical evaluation of large amounts of data, or processing and display of data by computer, aid in presentation and interpretation. Oftentimes, these methods rely on the same fundamental perspectives in presenting data.       

What if one changed the perspective? More than that, how about changing a really basic approach? In a previous installment of this column (15), we discussed the standard definition of the kilogram and the standard definition of the dalton. The consensus standards provide an underlying uniformity for our data and our discussions. With data processing, we can now easily shift standards and alter perspective. But 50 years ago, consider the value of such a new perspective (16):  

The problems of computing, storing, and retrieving precise masses of the many combinations of elements likely to occur in the mass spectra of organic compounds are considerable. They can be significantly reduced by the adoption of a mass scale in which the mass of the CH

 radical is taken as 14.0000 mass units. The advantage of this scale is that ions differing by one or more CH

 groups have the same mass defect. The precise masses of a series of alkyl naphthalene parent peaks, for example, are 127.9195, 141.91 95, 155.91 95, etc. Because of the identical mass defects, the similar origin of these peaks is recognizable without reference to tables of masses.   

That quote is from the abstract of a 1963 article (16) by Edward Kendrick, from the Analytical Research Division of Esso Research and Engineering. Kendrick confronted the reality that high resolution data (then measured at a resolving power of 5000) could be obtained for ions up to 

600 and concluded that "over one and a half million entries would be required to cover the ions up to mass 600. A mass scale with C

 = 14.00000 enables the same data — [that is], up to mass 600 — to be expressed in less than 100,000 entries. This reduction is a consequence of the same defect's being repeated at intervals of 14 mass units — [that is], one (CH

) group." It is not difficult to understand that a scientist at Esso would deal with a great many compounds that could differ by an integral number of methylene units (CH

). Given this challenge, the proposed approach and a shift in perspective was extraordinarily insightful. The eponymous term is the 

, and the difference between the standard mass scale and that proposed is called the 

. A mass unit called the kendrick (Ke) has also been used. 

The original publication by Kendrick (16) provides the rationale for this change in perspective and contains multiple tables used to illustrate its usefulness. A short summary is provided here. In the Kendrick perspective, the mass of the methylene (CH

) unit is set at exactly 14 Da. The International Union of Pure and Applied Chemistry (IUPAC) mass in daltons can be converted to the mass on the Kendrick scale by division by 1.0011178. The basis in the methylene unit can also be shifted to other functional groups, creating a mass scale linked to that group. The Kendrick mass defect is the difference between the nominal Kendrick mass and the exact Kendrick mass, in parallel to the IUPAC definitions. The value of this shift in perspective advocated by Kendrick is illustrated in the process called a 

. In this analysis, the Kendrick mass defect is plotted against the nominal Kendrick mass for ions in the mass spectrum (Figure 3). Successive members of a homologous series (differing in the number of methylene units) are represented by dots (representing ions observed in the mass spectrum) on the horizontal axis. After the composition of one ion is established, the composition of other ions can be established via the position on the plot. The Van Krevelen diagram, described in 1950, also adopted a perspective-shifting approach, but focused on other functional groups (17).  

Figure 3: A Kendrick mass analysis plot, with Kendrick mass defect on the y-axis, and Kendrick mass on the x-axis. Successive dots on an x-axis line represent ions observed (or not observed; a binary observation) from related compounds separated by a methylene unit (see red bar). Figure adapted from reference 22.			

Kendrick was confronting a daunting world of MS data recorded at a resolving power of 5000. We have advanced to the routine measurement of petroleomics data at 10–15× that resolving power (18) and extending to 10× higher mass, but the need for a meaningful display of the data persists. The Kendrick mass defect is just one of several "mass defects" that can inform our interpretation of mass spectrometric data (19). Increasingly, the methods of informatics give us new tools to approach the mass spectrometric data. The use of Kendrick mass data and Van Krevelen diagrams have been discussed within the larger context of high-precision frequency measurements recorded by many instrumental platforms, and the use of those data to discern complex molecular structures (20,21). 

We use eponymous terms in mass spectrometry because they efficiently convey information as well as crystallize and honor the achievements of an individual. Table I presents only a few such eponymous terms; many more may be in limited and specialized use. Whether they enter a more general use ultimately depends on the consensus of the community, which develops over years. Discovering the background of eponymous terms used in mass spectrometry provides a useful perspective of the development of the field. 

http://www.organic-chemistry.org/namedreactions/

http://www.chem.ox.ac.uk/vrchemistry/nor/

http://en.wikipedia.org/wiki/List_of_inventions_named_after_people

 (5) C.F. Brannock, as described in the New York Times of June 9, 2013. See 

http://www.brannock.com/cgi-bin/start.cgi/brannock/history.html

(13)  S. Wright, S. O'Prey, R.R.A. Syms, G. Hong, and A.S. Holmes, 

(18)  C.A. Hughey, C.L. Hendrickson, R.P. Rodgers, A.G. Marshall, and K. Qian, 

(20)  N. Hertkorn, C. Ruecker, M. Meringer, R. Gugisch, M. Frommberger, E.M. Perdue, M. Witt, and P. Schmitt-Koplin, 

(21)  T. Grinhut, D. Lansky, A. Gaspar, M. Hertkorn, P. Schmitt-Kopplin, Y. Hadar, and Y. Chen, 

https://en.wikipedia.org/wiki/Kendrick_mass

 doesn't have anything in mass spectrometry named after him or any pictures for sale on eBay. As a consolation prize, he does have beer, a theme park and gardens, and a company that markets vacuum equipment, all of which carry the name "Busch." None of these have any connection to the author, but the "Busch" neon sign purchased at the brewery gift shop sometimes provides a new perspective. This column is the sole responsibility of the author, who can be reached at 

Mass Spectrometry Forum

An important method of recognition in the scientific community is to use an individual's name in a description of the contribution, known as an eponymous tribute. Mass spectrometry showcases a number of inventions named after the inventor. Here, we explain two: the Brubaker prefilter and Kendrick mass.

Eponymous Mass Spectrometry: The Brubaker Prefilter and Kendrick Mass

Here we examine what is commonly called a Raman image and discuss how it is rendered. We consider a Raman image to be a rendering as a result of processing and interpreting the original hyperspectral data set. Building on the hyperspectral data rendering, we demonstrate the use of Raman imaging for the characterization of thin-film and ion-implanted silicon (Si) structures. A Raman image does not self-reveal solid-state structural effects and requires either the informed selection and assignation of the as-acquired hyperspectral data set by the spectroscopist or the use of statistical software applications. High spectral resolution allows us to clearly resolve the substrate Si Raman scattering from that of the polysilicon film. Consequently, high spectral resolution allows us to strongly resolve (structurally, not spatially) or contrast the substrate Si and polysilicon film in Raman images; that is, the spectral resolution contributes to the chemical or physical differentiation of spectrally similar materials in thin-film structures.

The primary goals of this column installment are to first examine in more detail what is commonly called a "Raman image" and to discuss how it is "rendered," and second to demonstrate the use of Raman imaging for the characterization of thin-film and ion-implanted silicon (Si) structures. The Raman images of Si and their rendering from the hyperspectral data sets presented here should encourage the reader to think more broadly of Raman imaging applications of semiconductors in electronic and other devices, such as microelectromechanical systems (MEMS). Raman imaging is particularly useful for revealing the spatial heterogeneity of solid-state structures in semiconductor devices. Here, test structures consisting of substrate silicon, silicon dioxide, polycrystalline silicon, and ion-implanted silicon are analyzed by Raman imaging to characterize the solid-state structure of the materials. 

Before we begin our discussion on Raman imaging of silicon structures, we need to carefully consider and understand what actually constitutes a "Raman image." Consider the black and white photograph, perhaps the most basic type of image with which we are familiar. One can think of it as a three-coordinate system, with two spatial coordinates and one brightness (independent of wavelength) coordinate. Progressing to a color photograph, we now have a four-coordinate system by adding a chromaticity coordinate to the original three of a black and white image. That chromaticity coordinate in the color photograph is the key to thinking about hyperspectral imaging in general and Raman imaging in particular. Whether color discrimination is due to the cone cells in our eyes, the wavelength sensitive dyes in color photographic film, or the color filters fabricated onto charge-coupled device (CCD) sensors, a wavelength selection is imposed on the image. That color discrimination may not faithfully render a color image that corresponds to the true color of the object; for example, people who are color blind may not see the true colors of an object. In our daily lives, it is the combination of shape (two spatial coordinates), brightness (intensity coordinate), and color (chromaticity coordinate) that we interpret in the images we see to draw meaning and respond to the objects from which they originate. In that sense, image interpretation comes naturally to us and we begin to do it from a very early age without giving it much analytical thought. 

When applying these same basic imaging principles of the four-coordinate system to hyperspectral imaging, the interpretation of the spectral (chromaticity) and intensity (brightness) coordinates are no longer "natural" and require mathematical thinking for proper interpretation. This mathematical thinking by the spectroscopist comes in the form of direct selection of spectral band position, width, shape, and resolution from closely spaced bands. The intensity coordinate is most often interpreted as the value at one particular peak or wavelength relative to that of another. However, even the simple application of an intensity coordinate requires the removal of background light unrelated to the spectroscopic phenomenon of interest. Furthermore, there are band-fitting operations and width and shape measurements that can be made and plotted against the two spatial coordinates to render an image. One can also apply any number of statistical and chemometric tools found in most scientific software. Therefore, whether the spectroscopist is directly engaged in the selection of spectral features and processing of the hyperspectral data set or simply uses statistical software operations, the spectral image is still the rendered product from a group of mathematical functions operating on the data. 

Now let's apply these hyperspectral imaging considerations to Raman imaging. We are all too familiar with the fluorescent background that often accompanies Raman scattering. In some instances, the spectroscopist will digitally subtract the fluorescent background before rendering a spectrum that consists almost exclusively of Raman data. By analogy, we might expect the same practice to hold if we wish to call an image a "Raman image" rather than a "spectral image" (one that includes data from all light from the object). For example, if the signal strength in a hyperspectral image detected at a particular Raman shift consists of 90% fluorescence and only 10% Raman scattering, it would not be correct to call that a Raman image. However, if the fluorescent background and any contribution from a light source other than Raman scattering is first removed, then we have a Raman image. Having done that, we are not necessarily finished rendering our Raman image. If we wish to spatially assign chemical identity or another physical characteristic to the image, the data must be interpreted either through a spectroscopist's understanding of chemistry and physics or through the statistical treatment of the data. The spectral image data as acquired are not self-selecting or -interpreting. It is only after applying the spectroscopist's judgment or statistical tools that one renders a color coded Raman image of compound A, compound B, and so on. 

I hope that you can see (pun intended) that Raman imaging is not an entirely simple matter or as intuitive as photography. Rather, it requires careful attention to the mathematical operation on the data and interpretation of the results. A Raman image is rendered as a result of processing and interpreting the original hyperspectral data set. The proper interpretation of the data to render a Raman image requires an understanding or at least some knowledge of the processes by which the image was generated. 

Imaging Strain and Microcrystallinity in Polysilicon

The development of Raman spectroscopy for the characterization of semiconductors began in earnest in the mid-1970s, and micro-Raman methods for spatially resolved semiconductor analysis were being used in the early 1980s. An account of the work in those early years can be found in the excellent book by Perkowitz (1). Perhaps even more surprising to young readers is the fact that one can find the words "Raman image" in either the title or text of some of those 1980s publications (2–4). Much of that early work involved the development of micro-Raman spectroscopy for the characterization of polycrystalline silicon (5–8), also known as polysilicon, which is still used extensively in fabricated electronic devices. Theoretical and experimental work was directed toward an understanding of how strain, microcrystallinity, and crystal lattice defects or disorder can all affect the Raman band shape, position, and scattering strength of single and polycrystalline silicon (9–11). The presence of nanocrystalline Si in polysilicon was confirmed through the combination of transmission electron microscopy and Raman spectroscopy, and the effects of extremely small Si grain dimensions on the Raman spectra were attributed to phonon confinement (12–15). As the crystalline grain size becomes smaller, comparable to the wavelength of the incident laser light or less, the Raman band broadens and shifts relative to that obtained from a crystalline domain significantly larger than the excitation wavelength. This has been explained in part by phonon confinement in which the location of the phonon becomes more certain as the grain size becomes smaller, and therefore the energy of the phonon measured must become less certain, consistent with the Heisenberg uncertainty principle. Now, with all that as historical background, let's take a fresh look at the Raman imaging of Si structures with work done and Raman images rendered in the first months of 2013. 

A collection of data from a polysilicon test structure is shown in Figure 1. A reflected white light image of the structure appears in the lower right-hand corner and a Raman image corresponding to the central structure in the reflected light image appears to its left. The plot on the upper left consists of all of the Raman spectra acquired over the image area and the upper right-hand plot is of the single spectrum associated with the cursor location in the Raman and reflected light images. The Raman data were acquired using 532-nm excitation and a 100× Olympus objective and by moving the stage in 200-nm increments over an approximate area of 25 μm × 25 μm. Now we make our first selection and operation on the hyperspectral data set. In this particular case, the Raman image is rendered through a color-coded plot of Raman signal strength over the corresponding color-bracketed Raman shift positions in the two upper traces.  

Figure 1: Raman hyperspectral data set from a polysilicon test structure with white reflected light image in lower right corner. The Raman image (lower left) corresponds to the color-coded bracketed regions in the spectral traces. The single spectrum (upper right) corresponds to the crosshairs in the Raman and white reflected light images.			

Let's discuss the reasoning behind our choice of the brackets and how they relate to differences in the solid-state structure of Si. If we were to have a merely elemental compositional image of this particular structure, we would expect it to be entirely uniform without spatial variation because Si would appear in every pixel of the image. However, if we distinguish the different solid-state structures of the Si by identifying pristine Si with the Brillouin zone center Raman band of 520.7 cm

  isolated in red brackets, microcrystalline and strained Si in green brackets, and a distribution of nanocrystallinity and strain in the blue brackets, we can render the Raman image in the lower left-hand corner. To some degree strain, microcrystallinity, and nanocrystallinity are commingled in the polysilicon regions of our images; however, we can make a crude distinction by attributing the blue regions as primarily a result of the nanocrystalline grain size. 

Now let's take a close look at what our rendering reveals in our Raman image, expanded for more detailed examination in Figure 2. The red regions consist of either substrate Si or grown single-crystal Si with different oxide thicknesses. The spatial variation of the single-crystal Raman signal strengths corresponds precisely with the physical optical effects of the oxide thicknesses and even small contaminants or defects seen in the reflected light image. Furthermore, because of the thinness of the polysilicon A alone, one can see through the green polysilicon A component to the underlying red substrate Si, particularly on the left and right of the upper and lower portions of Figure 2. The spatial variation of the microcrystalline or strained Si green component signal strength corresponds to both the central strip consisting of polysilicon B deposited on polysilicon A and the granular variation of polysilicon A seen in the reflected light images. Note that the polysilicon A bright green speckles in the Raman image correspond precisely to black speckles in the reflected light image. Careful examination of the central strip reveals these same speckles in the polysilicon A blurred somewhat by the polysilicon B deposited on top of it.  

Figure 2: The Raman image from Figure 1. Red corresponds to single crystal silicon, green is the strained and microcrystalline polysilicon, and blue is the nanocrystalline silicon. Polysilicon B (in the center) was deposited on the wider, underlying polysilicon A.			

Next let's turn our attention to the blue nanocrystalline portion of the image. The formation of nanocrystalline Si occurs almost exclusively in polysilicon A and only on top of the central single-crystal Si structure and not the substrate Si. Note that the nanocrystallinity occurs primarily along the edge of the polysilicon A, but disappears as the polysilicon A continues along the Si substrate. Also, we see that some of the polysilicon A speckles appear blue, thereby revealing nanocrystallinity over the central structure, but not over the Si substrate. 

In summary, the intensity variations of the single-crystal Si comport with the physical optical effects of varying oxide film thickness and surface contaminants. Also, this Raman image reveals the spatially varying nanocrystallinity, microcrystallinity, and strain in polysilicon. We can infer that these structural differences occur either as a result of processing conditions or from interactions with the adjacent or underlying materials in which the polysilicon is in contact. This exercise clearly demonstrates that a Raman image does not self-reveal solid-state structural effects without the informed selection and assignation of the as acquired hyperspectral data set. 

Imaging Crystal Lattice Damage Induced by Ion Implantation

Next, we examine the Raman imaging of that portion of our polysilicon test structure in which single-crystal silicon and polysilicon portions have been subjected to high energy ion implantation for the placement of dopants in the circuitry. The entire hyperspectral data set of one such region is shown in Figure 3. Again we have a structure consisting of Si substrate, isolated single-crystal Si, polysilicon A and B, silicon oxides, and a region implanted at high energy with As

. The Raman spectra were acquired over an area of approximately 30 μm × 30 μm using a 100× Olympus objective with 532-nm excitation and moving the electronically controlled stage in 200-nm increments.  

Figure 3: Raman hyperspectral data set from an ion-implanted polysilicon test structure with white reflected light image in the lower right corner. The Raman image (lower left) corresponds to the color-coded bracketed regions in the spectral traces. The single spectrum (upper right) corresponds to the lower left-hand corner of the Raman image. The ion-implanted Si yields only a weak Raman spectrum that resembles amorphous silicon, thereby appearing black in the Raman image. To get a better perspective on the importance of Raman imaging registration with the corresponding white light image see the YouTube video called "Raman Image Registration" (www.youtube.com/watch?v=rhHCOz-YEtc&feature=youtu.be).			

High energy ion implants disrupt the crystal lattice to varying degrees depending on the atomic mass, dose, and kinetic energy of the implanted ions. In this case, the heavy arsenic ions have completely damaged the crystal lattice. The long-range translational symmetry of the Si has been disrupted such that the Raman scattering from the ion-implanted Si arises from phonons throughout the Brillouin zone and not just the Brillouin zone center as in crystals. Consequently, ion-implanted Si yields only a weak Raman spectrum that resembles amorphous silicon (16–20). For this reason, the implant appears dark in the Raman image. Note that ion implantation affects the reflectance of the Si and that is why the implanted regions appear lighter than the adjacent unimplanted Si in the white reflected light image. Therefore, we have a good way of confirming the accuracy of our Raman image rendering by its registration with the white reflected light image. To get a better perspective on the importance of Raman imaging registration with the corresponding white light image see the YouTube video called "Raman Image Registration" (

www.youtube.com/watch?v=rhHCOz-YEtc&feature=youtu.be

Let's more carefully examine the expanded Raman image of the implanted structure shown in Figure 4. The substrate and isolated single-crystal Si can be seen in the lower half of the image and can also be seen underneath the polysilicon structures in the upper half. The spatial variations of the red intensities correspond well to the edges and contaminants or defects in the white reflected light image. The ion-implanted region contrasts strongly with single-crystal Si and polysilicon because of the weakness of the Raman signal from implanted, amorphized Si in all three of the bracketed spectral regions. Single-crystal Si and the lower edge of polysilicon A have both been implanted and the Raman image reveals the damage to the crystal lattice of both forms of Si. Also, note the sharpness of the Raman image and its registration with the reflected light image, which reveal the efficacy of Raman imaging as a means for ion implant placement and registration. To get a better perspective on how ion implantation affects the Raman image and to view individual spectra in the implant region see the YouTube video called "Raman Imaging of Implanted Si and Polysilicon" (

http://www.youtube.com/watch?v=Cf41DKhPzCs&feature=youtu.be

Figure 4: The Raman image from Figure 3. Red corresponds to single-crystal silicon, green corresponds to strained and microcrystalline polysilicon, and blue corresponds to nanocrystalline silicon. The ion-implanted Si yields only a weak Raman spectrum that resembles amorphous silicon, thereby appearing black in the Raman image. To get a better perspective on how ion implantation affects the Raman image and to view individual spectra in the implant region see the YouTube video called "Raman Imaging of Implanted Si and Polysilicon" (www.youtube.com/watch?v=Cf41DKhPzCs&feature=youtu.be).			

The polysilicon portions of the structure reveal the bright speckles from the polysilicon A that directly correspond to each of the black speckles in the white light reflected image. In contrast to the Si structures shown in Figures 1 and 2, the processes or adjacent materials have not induced the same degree of nanocrystallinity in this location of polysilicon A. Here, again, we have a structure with both forms of polysilicon and isolated Si, but the edges induced only a small amount of nanocrystallinity as revealed by the light blue streak along the polysilicon edge. Also, we are again able to "see" through the polysilicon to the Si substrate or isolated single-crystal Si. Note that the low energy limit of the red bracket begins at 520 cm

, and so the red region does reasonably differentiate substrate and isolated single-crystal Si from polysilicon. The Raman images in the next section allow us more insight into this effect.  

Imaging and the Importance of Spectral Resolution

Now we address the importance of spectral resolution for the generation of Raman images of thin-film structures in general and diffusely reflecting Si-based electronic devices in particular. Figure 5 shows the hyperspectral data set, the white reflected light image, and the approximately 14 μm × 22 μm Raman image from that portion of the test structure that consists only of polysilicon A and B on substrate Si. The L-shaped region of the images consists of (from bottom to top) substrate Si, polysilicon A, and polysilicon B. Note that the Raman bands of the different forms of Si in the hyperspectral data set (upper left) are better resolved than those in Figures 1–4. This is entirely because of differences in the solid-state structure of the polysilicon because the same spectrometer, grating (1800 gr/mm), and microscope objective were used for all data acquisition reported here. Raman spectra consisting wholly of broad, low energy bands peaking between 490 and 510 cm

  clearly reveal the presence of nanocrystallinity in the polysilicon.  

Figure 5: Raman hyperspectral data set from a polysilicon test structure with white reflected light image in lower right corner. The Raman image (lower left) corresponds to the color-coded bracketed regions in the spectral traces. The single spectrum (upper right) corresponds to the upper right-hand corner of the Raman image.			

Let's more carefully examine the material contrast in the expanded Raman image of the polysilicon structure shown in Figure 6. As we saw in the previous Raman images, polysilicon A shows a far greater propensity for forming nanocrystalline Si at the edges and distributed in some of the speckle. Polysilicon B also shows some nanocrystallinity at the edges; however, it is to a far lesser degree than in polysilicon A. The increased thickness of polysilicon B on top of A accounts for the bright green signal in the L-shaped region and the very dim red contribution from the underlying Si substrate. In that region, with only one layer of polysilicon A covering the substrate Si, the underlying red single-crystal Si signal emerges strongly through that of the green polysilicon A. The high spectral resolution allows us to clearly resolve the single-crystal Si Raman bands in the red bracket from those of the polysilicon in the green bracket. Consequently, in this particular structure of only polysilicon on Si substrate, the high spectral resolution allows us to strongly resolve (structurally, not spatially) or contrast the substrate Si and polysilicon in Raman images; that is, the spectral resolution contributes to the chemical or physical differentiation of spectrally similar materials in thin-film structures.  

Figure 6: The Raman image from Figure 5. Red corresponds to substrate silicon, green corresponds to strained and microcrystalline polysilicon, and blue corresponds to nanocrystalline silicon.			

Excitation wavelength also plays an important role in depth profiling semiconductor structures (19). An excitation wavelength shorter than 532 nm would have a shallower depth of penetration in Si (21), and so we might expect the Raman images generated using different excitation wavelengths to look different. At some shorter excitation wavelength, depending on the thickness of the polysilicon, one would no longer be able to detect the underlying Si substrate or single-crystal Si because of the shallow depth of penetration. We'll have to save Raman image depth profiling by excitation wavelength selection for another installment of "Molecular Spectroscopy Workbench." 

We have examined in more detail what is commonly called a Raman image and discussed how it is rendered. We claim that a Raman image is rendered as a result of processing and interpreting the original hyperspectral data set. The proper interpretation of the data to render a Raman image requires an understanding or at least some knowledge of the processes by which the image was generated. Building on the hyperspectral data rendering, we demonstrate the use of Raman imaging for the characterization of thin-film and ion-implanted silicon structures. We show that a Raman image does not self-reveal solid-state structural effects, but requires either the informed selection and assignation of the as acquired hyperspectral data set by the spectroscopist or the use of statistical software applications. High spectral resolution allows us to clearly resolve the substrate Si Raman scattering from the polysilicon film. Consequently, high spectral resolution allows us to strongly resolve (structurally, not spatially) or contrast the substrate Si and polysilicon film in Raman images; that is, the spectral resolution contributes to the chemical or physical differentiation of spectrally similar materials in thin-film structures. 

Optical Characterization of Semiconductors: Infrared, Raman, and Photoluminescence Spectroscopy

 (Academic Press,  London, UK, 1993), Chap. 6. 

 (2) K. Mizoguchi, Y. Yamauchi, H. Harima, S. Nakashima, T. Ipposhi, and Y. Inoue,

 (3) S. Nakashima, K. Mizoguchi, Y. Inoue, M. Miyauchi, A. Mitsuishi, T. Nishimura, and Y. Akasaka, 

 (4) K. Kitahara, A. Moritani, A. Hara, and M. Okabe, 

 (5) Y. Inoue, S. Nakashima, A. Mitsuishi, T. Nishimura, and Y. Akasaka, 

 (6) D.R. Tallant, T.J. Headley, J.W. Medernach, and F. Geyling, 

 (8) G. Harbeke, L. Krausbauer, E.F. Steigmeier, A.E. Widmer, H.F. Kappert, and G. Neugebauer,

 (9) G.-X. Cheng, H. Xia, K.-J. Chen, W. Zhang, and X.-K. Zhang, 

(13) V.A. Volodin, M.D. Efremov, and V.A. Gritsenko, 

(14) Y. He, C. Yin, G. Cheng, L. Wang, X. Liu, and G.Y. Hu, 

(15) H. Xia, Y.L. He, L.C. Wang, W. Zhang, X.N. Liu, X.K. Zhang, D. Feng, and H.E. Jackson,

(16) R. Tripathi, S. Kar, and H.D. Bist, 

Molecular Spectroscopy Workbench

What exactly is a "Raman image" and how is it rendered? The authors explain those points, and demonstrate the use of Raman imaging for the characterization of thin-film and ion-implanted silicon structures. High spectral resolution makes it possible to resolve or contrast the substrate silicon and polysilicon film in Raman images and thus aids in the chemical or physical differentiation of spectrally similar materials.

Raman Imaging of Silicon Structures

Both Raman spectroscopy and surface enhanced Raman spectroscopy (SERS) are proving to be invaluable tools in the field of biomedical research and clinical diagnostics. The robust, compact, fit-for-purpose Raman spectrometer designs are appropriate for use in surgical procedures to help surgeons assess tumors and allow rapid decisions to be made. Raman systems are also being developed for molecular diagnostic testing to detect and measure human cancer biomarkers. Based on the SERS technique, this approach potentially could change the way bioassays are performed to improve both the sensitivity and reliability of testing. The two applications highlighted in this review, together with other examples of the use of Raman spectrometry in biomedical research areas such as the identification of bacterial infections, are clearly going to make the technique an important part of the medical toolbox, as we continually strive to improve diagnostic techniques and bring a better health care system to patients.

In recent years, Raman spectroscopy has gained widespread acceptance in applications that span from the rapid identification of unknown components to detailed characterization of materials and biological samples. The technique's breadth of application is too wide to reference here, but examples include quality control (QC) testing and verification of high purity chemicals and raw materials in the pharmaceuticals and food industries (1–3), investigation of counterfeit drugs (4), classification of polymers and plastics (5), characterization of tumors, and the detection of molecular biomarkers in disease diagnostics (6–8) and theranostics (9).  

One of the most exciting application areas is in the biomedical sciences. The major reason behind this surge of interest is that Raman spectroscopy is an ideal technique for molecular fingerprinting and is sensitive to the chemical changes associated with disease. Furthermore, components in the tissue matrix, principally those associated with water and bodily fluids, give a weak Raman response, thereby improving sensitivity for those changes associated with a diseased state (10). The growing body of knowledge in our understanding of the science of disease diagnostics and improvements in the technology has led to the development of fit-for-purpose Raman systems designed for use in surgical theaters and doctors' surgeries, without the need for sending samples to the pathology laboratory (8).  

Surface-enhanced Raman spectroscopy (SERS) is a more sensitive version of Raman spectroscopy, relying on the principle that Raman scattering is enhanced by several orders of magnitude when the sample is deposited onto a roughened metallic surface. The benefit of SERS for these types of applications is that it provides well-defined, distinct spectral information, enabling characterization of various states of a disease to be detected at much lower levels. Some of the many applications of Raman and SERS for biomedical monitoring include   					

Cytology investigations at the cellular level

Direct investigation of cancerous tissues

Surgical targets and treatment monitoring 

The basics of Raman spectroscopy are well covered elsewhere in the literature (11). However, before we present some typical examples of both Raman and SERS applications in the field of cancer diagnostics, let's take a closer look at the fundamental principles and advantages of SERS. 

Principles of Surface-Enhanced Raman Spectroscopy

The principles of SERS are based on amplifying the Raman scattering using metal surfaces (usually Ag, Au, or Cu), which have a nanoscale roughness with features of dimensions 20–300 nm. The observed enhancement of up to 10

-fold can be attributed to both chemical and electromagnetic effects. The chemical component is based on the formation of a charge-transfer state between the metal and the adsorbed scattering component in the sample and contributes about three orders of magnitude to the overall enhancement. The remainder of the signal improvement is generated by an electromagnetic effect from the collective oscillation of excited electrons that results when a metal is irradiated with light. This process, known as the surface plasmon resonance (SPR) effect, has a wavelength dependence related to the roughness and atomic structure of the metallic surfaces and the size and shape of the nanoparticles. For a more detailed discussion of SERS and its applications, see reference 12.    

Detection by SERS has several benefits. Firstly, fluorescence is inherently quenched for an adsorbate analyte on a SERS surface resulting in fluorescence-free SERS spectra. This is primarily because the Raman signal is enhanced by the metal surface, and the fluorescence signal is not. As a result, the relative intensity of the fluorescence is significantly lower than the Raman signal and in many occasions is not observable. Secondly, signal averaging can be extended to increase sensitivity and lower detection levels. Finally, SERS spectral bands are very sharp and well-defined, which improves data quality, interpretability, and masking by interferents. Recent work using silver nanoparticles as the enhancing substrate has shown that SERS can be applied to single-molecule detection, rivaling the performance of fluorescence measurements (13). 

Although SERS has many advantages it also has several disadvantages that are worth noting. The first is that unlike traditional Raman spectroscopy, SERS requires sample preparation. The ability to measure samples in vivo without the need for sample pretreatment is one of the major advantages of traditional Raman spectroscopy. For that reason, it is important to understand the signal enhancement benefits of SERS compared to the ease of sampling of traditional Raman spectroscopy when deciding which technique to use (14). Second, high performance SERS substrates are difficult to manufacture and, as a result, there is typically a high degree of spatial nonuniformity with respect to the signal intensity (15). For example, the SERS substrates used in the research being carried out by the University of Utah (described later in this article) tend to have much higher concentrations of analyte on the edges of the sampling area than in the center. Lastly, it is important to note that when the sample is deposited onto the surface, the molecular structure (the nature of the analyte) can be altered slightly resulting in differences between traditional Raman spectra and SERS spectra (16). However, it should be emphasized that there have been improvements in the quality of substrates being manufactured today and new materials such as Klarite (Renishaw Diagnostics) are showing a great deal of promise in terms of consistency and performance (17). 

Raman Spectroscopy Applied to Biomedical Studies

There are several important functional groups related to biomedical testing that have characteristic Raman frequencies. Tissue samples include components such as lipids, fatty acids, and protein, all of which have vibrations in the Raman spectrum. The most significant spectral regions include 					

X-H bonds (for example, C-H stretches): 4000–2500 cm

Multiple bonds (such as N≡C): 2500–2000 cm

Double bonds (for example, C=C, N=C): 2000–1500 cm

Complex patterns (such as C-O, C-N, and bands in the fingerprint region): 1500–600 cm

The identification and measurement of Raman scattering in these spectral regions can be applied to the following biomedical applications: 					

Measurement of biological macromolecules such as nucleic acids, proteins, lipids, and fats

Investigation of blood disorders such as anemia, leukemia, and thalassemias (inherited blood disorder)

Detection and diagnosis of various cancers including brain, breast, pancreatic, cervical, and colon 

Assaying of biomarkers in studying human diseases

Understanding cell growth in bacteria, phytoplankton, viruses, and other microorganisms

Analysis of abnormalities in tissue samples such as brain, arteries, breast, bone, cervix, embryo media, esophagus, gastrointestinal tract, and the prostate gland

Some typical examples of Raman spectra of biological molecules are shown in Figure 1. Figure 2 gives a summary of some common biochemical functional groups with their molecular vibrational modes and frequency ranges, which are observed in compounds such as proteins, carbohydrates, fats, lipids, and amino acids (6,8,10,18–20). Identification of these functional groups can be used as a qualitative or quantitative assessment of a change or anomaly in a sample under investigation.  

 Figure 1: Raman spectra of various biological compounds. Adapted from reference 21.			

Let's take a look at two examples of the use of both conventional Raman spectroscopy and SERS, specifically for cancer diagnostic purposes. The first example uses traditional Raman spectroscopy for the assessment of breast cancer and the second example takes a look at SERS for the screening of pancreatic cancer biomarkers.  

Figure 2: Some common functional groups with their molecular vibrational modes and frequency ranges observed in various biochemical compounds. Adapted from reference 18.			

Detection and Diagnosis of Breast Cancer 

The surgical management of breast cancer for some patients involves two or more operations before a surgeon can be assured of removing all cancerous tissue. The first operation removes the visible breast cancer and samples the lymph nodes for signs of the disease spreading. Intraoperative methods for testing the health of the lymph nodes involve classical tissue mapping techniques such as touch imprint cytology and histopathological, frozen section microscopy. The problem with both of these methods is that they have poor sensitivity and require an experienced pathologist to interpret the data and relay the "benign" or "malignant" diagnostic report to the surgeon. For that reason, these intraoperative methods have not been widely adopted by the medical community and, as a result, most procedures still require samples to be sent to a laboratory for further testing. If tests confirm that the cancer has spread, then the patient must undergo an additional surgery to remove all the nodes in the affected area.  

A preferable approach would be for an intraoperative assessment, which would allow for the immediate excision of all of the axillary lymph nodes, if required. Studies by researchers at Cranfield University and Gloucestershire Royal Hospital in the United Kingdom have shown that Raman spectroscopy can detect differences in tissue composition, particularly a wide range of cancer pathologies (22). This group used a portable Raman spectrometer with a 785-nm excitation laser wavelength and a fiber-optic probe (MiniRam, B&W Tek) and principal component analysis (PCA) along with linear discriminant analysis (LDA) data processing to evaluate the Raman spectra of lymph nodes. They confirmed that the technique can match the sensitivities and specificities of histopathological techniques that are currently being used.  

The initial work in this study involved taking Raman spectra of the same samples that were being evaluated by the traditional intraoperative techniques. The study, which assessed 38 (25 negative and 13 positive) axillary lymph nodes from 20 patients undergoing surgery for newly diagnosed breast cancer, compared the results with a standard histopathological assessment of each node. The results showed a specificity of 100% and sensitivity of 92% when differentiating between normal and metastatic nodes. These results are an improvement on alternative intraoperative approaches and reduce the likelihood of false positive or false negative assessment (20). Typical Raman spectra of lymph nodes from benign breast tissue and malignant tumors are shown in Figure 3. It can be seen that while there are very subtle differences between the two spectra, the data classes can be discriminated using multivariate classification tools such as PCA-LDA and support vector machine (SVM) classification applied over the 800–1800 cm

 spectral range. By using such supervised classification tools, Raman spectroscopy is sensitive enough to pick up the differences in the chemistry of the diseased state compared to that of the healthy tissue. For example, in Figure 3, subtle differences in the fatty acid peaks at 1300 and 1435 cm

, are highlighted from the loadings plots from PCA analysis of the data (19).  

Figure 3: Raman spectral peaks of lymph nodes of benign breast tissue (blue), together with a malignant breast tumor (red). Adapted from reference 21.			

More work is currently being done to support these findings at the University of Exeter and other research facilities. If confirmed, the next stage is to introduce the probe-based Raman spectrometer into an operating theater, where it can be used to assess the node as soon as it is removed from the patient.  

Additionally, many other groups are working on different approaches to cancer screening using Raman spectroscopy. Recent studies at the Massachusetts Institute of Technology (MIT) have shown that this technique has the potential to be built in to a biopsy needle, allowing doctors to instantaneously identify malignant or benign growths before an operation (23). Recent studies have indicated that the efficacy of these techniques can be enhanced by including a wider spectral region beyond 1800 cm

  in the model to increase specificity (24). Finally, for the characterization of highly fluorescent biological tissues, such as liver and kidney samples, researchers have shown evidence to suggest that shifting the excitation laser wavelength to 1064 nm provides a cleaner Raman response with reduced fluorescence compared to excitation at 785 nm (25). Dispersive Raman technology at 1064 nm is relatively new and is slowly gaining traction. Most biological tissues produce fluorescence to a greater or lesser extent, so reducing this effect will improve the Raman signal and potentially improve the analysis and the quality or accuracy of the diagnostic outcome. 

SERS Immunoassay for Pancreatic Cancer Biomarker Screening

Pancreatic adenocarcinoma is the fourth most common cause of cancer deaths in the United States, with a 5 year survival rate of ~6% and an average survival rate of <6 months. The reason for this high ranking is that the disease can only be detected and diagnosed by conventional radiological and histopathological detection methods when it has progressed to an advanced stage. After it has been diagnosed, resection (partial removal) of the pancreas is the normal treatment (26).  

There are potentially more than 150 biomarkers found in serum for the detection of pancreatic and other cancers. One of the most prevalent is carbohydrate antigen 19-9 (CA 19-9), a mucin type glycoprotein sialosyl Lewis antigen (SLA), which shows elevated levels in ~75% of patients with pancreatic adenocarcinoma (27). The most common diagnostic test for these biomolecular compounds is enzyme-linked immunosorbent assay (ELISA), which is a well-established bioassay that uses a solid-phase enzyme immunoassay to detect the presence of an antigen in serum samples. It works by attaching the sample antigen to the surface of the sorbent. Then a specific monoclonal antibody, which is linked to the enzyme, is applied over the surface so it can bind to the antigen. In the final step, a substance containing the enzyme's substrate is added, which generates a reaction to produce a color change in the substrate (28). Even though this technique is widely used for bioassays, it does have some limitations with regard to the detection of cancer biomarkers. For example, early stage markers are present at levels well below current detection capabilities. In addition, 100–200 μL of sample is typically required to achieve a low enough limit of detection. Moreover, CA 19-9 is unlikely to be detected in patients with tumors smaller than 2 cm. It is further complicated by the fact that an estimated 15% of the population cannot even synthesize CA 19-9. The limitations in the ELISA approach (29) have led to the investigation of alternative methods, including an immunoassay based on SERS. This offers the exciting possibility of an alternative, faster way of detecting many different biomarkers in the same assay.  

A portable Raman spectrometer (MiniRam, B&W Tek) was used in a recent study at the University of Utah's Nano Institute to detect and quantify the CA 19-9 antigen in human serum samples (30). It showed the advantages of generating a SERS-derived signal from a Raman reporter molecule attached to the biomolecule marker analyte using gold nanoparticles. This process is carried out using an analyte sandwich immunoassay in which monoclonal antibodies are covalently bound to a solid gold substrate to specifically capture the antigen from the sample. The captured antigen is in turn bound to the corresponding detection antibody. This complex is joined together with gold nanoparticles that are labeled with different reporter molecules, which serve as extrinsic Raman labels (ERLs) for each type of antibody. The presence of specific antigens is confirmed by detecting and measuring the characteristic SERS spectrum of the reporter molecule. Let's take a closer look at how this works in practice for these biomarkers and how the sample is prepared and read-out, in particular.   

The first step is the preparation of the ERLs. This is done by coating a 60-nm gold nanoparticle with the Raman reporter molecule, which in this case is 5,5'-dithiobis(succinimidyl-2- nitrobenzoate (DSNB). This complex is then coated with a target or tracer antibody using 

-hydroxysuccinimide (NHS) coupling chemistry. The DSNB serves two purposes: The nitrobenzoate group is the reporter molecule and the NHS group enables the protein to be coupled to the nanoparticle surface. 

The second step is to prepare the substrate for capture. This is done by resistively evaporating solid gold under vacuum using thin film deposition on a glass microscope slide and then growing a monolayer of dithiobis(succinimidyl propionate) (DSP) on the surface of the slide to link the capture antibody to the gold surface. Next, the appropriate antibody is attached to the substrate, based on the biomarker of interest. 

The third step is to make the immunoassay sandwich. This step involves incubating the slide with serum containing the CA 19-9 or MMP-7 antigens, which are then captured by surface-bound monoclonal antibodies and labeled with the ERLs. 

The fourth and final step is to read out the Raman signal of the functional group of interest, using a spot size of ~85 μm with a laser excitation wavelength of 785 nm. In the study, the symmetric nitro stretch band of the DNSB molecule at 1336 cm

  was used for quantitation by measuring the peak height to construct a calibration curve for various concentrations of the antigens spiked into serum. Real-world serum samples are then quantified using this calibration curve.  

Peer-Reviewed Article

Raman spectroscopy and surface-enhanced Raman spectroscopy (SERS) are proving to be valuable tools in biomedical research and clinical diagnostics. This review looks at two examples: the use of traditional Raman spectroscopy for the assessment of breast cancer and the use of SERS for the screening of pancreatic cancer biomarkers.

Spectroscopy-09-01-2013

Why Do We Need Quality Agreements?

Market Profile: Raman Spectroscopy in the Pharmaceutical Industry

Eponymous Mass Spectrometry: The Brubaker Prefilter and Kendrick Mass

Raman Imaging of Silicon Structures

The Use of Raman Spectroscopy in Cancer Diagnostics

Vol 28 No 9 Spectroscopy September 2013 Regular Issue PDF