Raman spectroscopy is a well-known technique used to identify materials and chemicals. Until the 1980s, most Raman instruments used dedicated monochromators with photomultiplier tube (PMT) detectors.1 These early systems enjoyed the high sensitivity that a PMT provides but lacked versatility and red response (700- to 1000-nm region). Today the Raman market is dominated by CCD detectors with versatile grating and high sensitivity in the red and near-IR regions. This article will explain how to select a CCD detector for Raman spectroscopy, focusing primarily on relevant CCD device parameters, their trade-offs and why they are important to Raman spectroscopy. The most recent trend is making small, portable handheld Raman instruments to measure such things as biohazardous materials, diseases and bacteria. The demand for high-performance sensors and instrumentation in this application is driving the evolution of new product development. Instrument manufacturers require image sensors such as CCDs to have low noise, high sensitivity, low etaloning and excellent linearity characteristics. Noise Because this new generation of handheld Raman instruments is often battery-powered, cooling the image sensor is not a viable solution for noise reduction. There are only a handful of CCD manufacturers in the world capable of producing noncooled image sensors with intrinsically low noise. Manufacturers who own a foundry tend to have better quality control over the CCD manufacturing process compared with those who outsource. Having such a foundry enables optimizing the wafer processing conditions to reduce the dark current. This is very important for Raman spectroscopy. Speed The topic of operating speed can be viewed from two perspectives: instrument and CCD sensor readout. The two topics can be confusing because the terminology for each opposes the other. And because the optical signal impinging on the CCD surface is extremely weak, people generally are not concerned about fast CCD readout speed. Reading out the CCD sensor in a matter of a few milliseconds to tens of milliseconds is perfectly acceptable for most applications. The tendency is to increase the integration time – on the order of a fraction of a second to several seconds – to improve the signal-to-noise ratio. Operating the CCD at high pixel frequencies, or high readout speed, will be counterproductive in terms of increased readout noise, increased power consumption and self-heating of the CCD chip, thus increasing the CCD dark signal. The rate at which Raman spectral data is acquired is very important for some applications, as stated earlier. If all other variables are held constant, spectral data can be acquired at twice the speed by using a back-thinned CCD with double the quantum efficiency of conventional CCDs. This means that a handheld instrument can perform material analyses faster, an important benefit for security applications. Quantum efficiency Raman signal – number of photons – typically is very weak, and detectors must capture every available photon to maximize the signal-to-noise ratio. The optical signal for a classical Raman measurement – not surface-enhanced – is extremely low: Only 0.001 percent of the excitation source is converted into Raman signal. This means that the Raman measurement system must be capable of detecting weak optical signals, down to the level of tens of photons. Figure 1. Quantum efficiency curve of a conventional back-thinned CCD (BT-CCD). Images courtesy of Hamamatsu Corp. Ta = ambient temperature. Back-thinned CCDs (BT-CCDs) are well suited for such low light detection because their quantum efficiency (QE) reaches 90 percent at the peak wavelength (Figure 1). Such devices have high QE in the UV and visible regions because the incident light interacts directly with the sensor’s active region. Making their transition from R&D to mass production, near-IR-enhanced BT-CCDs also are proving to be a benefit for Raman spectroscopy. Laser treatment of the BT-CCD can further increase the QE in the red and near-IR regions (Figure 2). The advantage of higher QE is improved signal-to-noise ratio. With this new technology, instrument designers can double the signal-to-noise ratio at 800 nm. Figure 2. Quantum efficiency curves of a front-illuminated CCD, BT-CCD (conventional type) and IR-enhanced BT-CCD. Etaloning The drawback to a back-thinned CCD is etaloning. As light passes through the thinned region (10 to 20 μm thick), a small percentage of the light hits structures, or layers, near the front side and reflects back into the active region, generating electrons (signal). This is particularly noticeable in the red and near-IR regions. The reflected signal creates a fringe pattern in the spectral response (Figure 3). The best technique to reduce etaloning is to create irregular patterns at the origin of reflection, preventing the light from reflecting back into the active region. By employing such patterns, Hamamatsu can reduce the etaloning by a factor of five. Reducing etaloning is of great benefit to the end user because the interference pattern would otherwise mask the actual spectral signature of the sample, making identification of the sample impossible. Figure 3. Comparison of etaloning in a conventional BT-CCD and a BT-CCD with low etaloning. Linearity Some applications of Raman spectroscopy require the ability to measure concentrations of samples. In such cases, it is necessary for the Raman instrument to have good linearity. Linearity of the CCD is governed by the design of the on-chip floating diffusion amplifier. Linearity often is not written as a specification, so be sure to contact CCD manufacturers and request the data. As a common practice, many instrument designers will write algorithms to compensate for nonlinearity. The task of linearity correction is made easier if you select a CCD that inherently has a good linear response. Electronic circuitry Equally important is the electronics design of the CCD readout circuitry and clock drivers. The highly optimized CCD sensors described previously are of no value without proper readout and interface electronics. Electronics designers have the option to use discrete components – amplifiers, correlated double sampling and analog-to-digital converters – or an “all in one” approach with an analog front-end (AFE) device. The trade-off is simple. AFEs offer simple implementation with low power consumption and compact size. On the other hand, discrete circuit designs offer the lowest readout noise (as much as 30 percent lower than an AFE design), but are more complicated to implement and often require an experienced low-noise analog design engineer. Handheld instrument designers must reduce physical size and electrical power consumption, so using an AFE will help accomplish this goal. There are many manufacturers of AFEs, allowing you to select a device that is compatible with your CCD detector. CCD biasing Today’s scientific-grade CCD requires multiple bias and clock voltages. A traditional circuit with CCD substrate grounded (0 V) will require DC power supplies ranging from —8 to 24 V. Grounding in these configurations is easier because the CCD substrate is common with earth ground. In this case, the CCD clock voltage will swing between —8 and 6 V. However, such a biasing scheme is difficult to create with a single input supply of 5 V/500 mA (USB bus power). Another approach is to float the CCD substrate (biased at 8 V). This will reduce the bias voltages and eliminate the need for bipolar, or negative, supplies. With the CCD substrate at 8 V, the clock voltage swinging from 0 to 14 V will have the same effect as the —6- to 8-V configuration. Look for common high-speed dual channel power MOSFET (metal-oxide semiconductor field-effect transistor) drivers on the market, or contact your CCD supplier for specific recommendations to accommodate your application. Optics Besides a detector, most Raman instruments incorporate high-spectral-response optics – a spectrometer – to distinguish spectral features of the sample (Figure 4). A transmission mode grating with a fiber-coupled entrance slit enables designers to achieve high optical throughput and narrow optical resolution. Figure 4. Block diagram of a spectrometer. However, everything is a trade-off; such is the case with pixel size, slit width and optical resolution. The larger the spectrometer entrance slit, the more photons enter the optical path, but the lower the achievable spectral resolution. Conclusion With the advances in sensor technology, designers can build compact instruments for Raman spectroscopy that have excellent sensitivity, speed and stability. When selecting a CCD sensor, designers must seek sensors with natively low dark current. In some applications, the use of a low-power thermoelectric cooler is necessary to reduce and stabilize the dark signal from the CCD. Because every photon counts, select a CCD with high quantum efficiency in the region of interest. An excellent choice is an enhanced back-thinned CCD because it offers the highest QE, low dark current and low readout noise. Meet the author John Gilmore is an image sensor marketing manager at Hamamatsu Corp. in Bridgewater, N.J.; e-mail: jgilmore@hamamatsu.com. Reference 1. Douglas Skoog et al (1998). Principles of instrumental analysis, 5th edition. Thomson Learning Inc., pp. 436-437.