Spectral Imaging

How to Separate the Colors

December 09, 2011

The goal of multiparameter fluorescence imaging is to collect photons emitted by the various fluorochromes into separate detection channels. For this purpose, it is necessary to spatially isolate fractions of the full emission spectrum, i.e. direct these fractions to differing directions. Classically, this separation is performed by “secondary dichroic mirrors” (referring to the main beam splitter which separates illumination from emission as “primary beam splitter”). Unexpectedly, it is also possible to use prisms (first choice) or gratings. This pristine true color separation by physically sorting the photons according to their color may be supplemented with mathematical unmixing if the primary separation was insufficient.

Secondary dichroic mirrors

Since about 1970, multiparameter fluorescence microscopy has been increasingly in demand in the field of biological microscopy. In the simplest case, the filters and dichroics in use for a single-channel setup will inherently allow recording of e.g. green and yellow/orange fluorescence upon blue or UV excitation (a meanwhile forgotten standard at that time was Feulgen staining, which could discern DNA and RNA by different emission colors). Widefield microscopy responds to this requirement by using a color camera. Sometimes, especially for quantitative measurements, the image is split and two channels are imaged in parallel on the same chip. Most advanced is the simultaneous use of two or more cameras in parallel. In true confocal scanning microscopy, a split of the sensing target is impossible. As a consequence, multiparameter fluorescence was immediately implemented by adding a second (third…) photomultiplier. The data were recorded in parallel and either directly displayed on the screen, e.g. in the 3 color channels of the monitor, or stored electronically for later analysis. To display more than 3 channels, the information must be distributed among the 3 available monitor channels, inevitably causing loss in separation and intensity resolution. Nevertheless, modern microscopy is not just about brilliant images (we can discern only three channels with our eyes anyway), but targets quantitative measurements. Here, any number of channels makes sense as long as the number of channels does not exceed the number of fluorochromic species in the sample.

Fig. 1: Layout of a 4-channel true confocal scanning microscope with secondary dichroic mirrors (Leica TCS NT, 1995). L1 and L2 are designed as “filter wheels”, whereas M is a fixed mirror (for 3-channel systems) or a fixed dichroic mirror (in 4-channel systems).
Sketch of the Leica TCS NT confocal microscope scanhead, as used back in 1995. The emission splitting was performed with 2 wheels holding a series of dichroic mirrors and a fixed (dichroic) mirror for the 4th channel.

The most obvious approach for directing the emission from fluorochromes emitting different colors to the set of sensors is the use of dichroic mirrors. A dichroic mirror will reflect light of shorter wavelength than the dichroic-specifying wavelength λ0 and transmit all colors of longer wavelength. This is true for a “longpass dichroic mirror”. A “shortpass dichroic mirror” will transmit the shorter wavelength and reflect the redder part of the spectrum. In Figure 2, a set of three secondary dichroics S1, S2 and S3 serve for splitting the full spectrum into 4 different directions, where the sensors can subsequently collect 4 different channels. The secondary dichroic mirrors have to be selected according to the spectral emission characteristics of the fluorochromes in use. As a consequence, a given set may fit to a number of similarly emitting fluorochromes, but fails if the fluorochrome combination has significantly different emission characteristics. To solve this problem, systems that use secondary dichroics for color separation are equipped with wheels or sliders in each split position. These are armed with a series of various dichroic mirrors that allow a (finite) number of different color bands to pass to the sensors. Obviously, this solution is not very flexible and requires a large amount of servo-technique and adjustment (that is expected to stay stable over many months,  at least. If many laser lines are installed, the number of potential emission bands will increase – and so the number of secondary dichroics required. With a white light laser source, the emission filter concept will fail to adapt sensibly and only a continuously tunable device will fit. Figure 1 shows an implementation of the secondary dichroic concept from 1995.

Fig. 2: Color dispersion by dichroic mirrors. A primary beam splitter P separates excitation from emission. The emission is subsequently split into four directions (this example) by secondary dichroics S1, S2 and S3. A regular dichroic mirror reflects short wavelength and transmits “redder” colors, although an inverse characteristic is equally possible.
Classical beam splitting for multichannel fluorescence recording is performed with secondary beam splitters. This concept has been employed for confocal systems Leica CLSM, Leica 4D and Leica NT. Modern confocal systems use a prism for dispersion, as in a

Dispersion by prism

Fig. 3: Sketch of a prism-based confocal detection module featuring 5 channels with individually tunable bands for emission collection.<br>1) Prism, 2) Sliders, 3) Detector
Dispersion of colors from a mixed source by a prism. Here, a sample is illuminated by a laser beam in a true confocal scanning microscope. The colored emission is fed through a prism, which disperses the directions according to the wavelength. A sketch of

The oldest method of (deliberately) separating colors of light was described by Sir Isaak Newton in his book “Opticks” of 1704: the employment of a prism. A copy of his drawing from that book is shown with the preface in this article. Today, our interpretation is that light of shorter wavelength will be diffracted at a boundary of optically different media more strongly than light of longer wavelength (to make a simple conclusion). If a mixture of colors – like the composed emissions of a set of fluorochromes – is fed through a prism, the composed emissions will be decomposed spectrally.

The strength of decomposition depends on many technical parameters, but is independent of the sample or the sensor. This is a very efficient and straightforward solution for the problem of having emissions from a set of fluorochromes pointing in different directions, where they can be subsequently recorded. In the simplest case, one would just place a series of detectors along the spectrum. This concept is realized, but has severe flaws in collection efficiency and flexibility. A better solution is a multiband device which allows individual selection of any fraction of the full spectrum for each sensor to be recorded.

A prism has the advantage of white (flat) transmission, i.e. there is no absorption modulation by the prism (within the specified spectral range). The transmission and the dispersion are independent of the direction of polarization. This is an important fact, as the emission of a fluorochrome is always unpolarized. And last but not least, the dispersion occurs only in one single direction – there are no other “orders” that may reduce the intensity in the selected order.

There is some discussion about the linearity of the dispersed spectrum. A prism-based spectrum is not linear with respect to the wavelength. For technical designs, this is not a problem, as long as one is not bound to use linear detector arrays, such as multi-anode photomutipliers or similar devices.

Fig. 4: Color dispersion by a prism. Due to the fact that the refractive power of a glass prism depends on the wavelength of the diffracted light, a composition of different wavelengths will be dispersed. In the case of white light, the dispersion pattern is a continuous spectrum. The emission of a multiply labeled fluorescence sample will yield a spectrum that has maxima at the emission maxima of the fluorochromes. The art is to collect as much as possible of a single dye by a single sensor.
Dispersion of colors from a mixed source by a prism. Here, a sample is illuminated by a laser beam in a true confocal scanning microscope. The colored emission is fed through a prism, which disperses the directions according to the wavelength. A sketch of

Dispersion by grating

Fig. 5: Comparison of efficiencies for gratings vs. prisms. The grating (blue curves) is optimized for a “blaze” wavelength and efficiency drops on both sides of that wavelength. Furthermore, the efficiency is good (at the blaze) for one polarization direction. The perpendicular polarization drops significantly within a short band. The efficiency of a prism depends mainly on the glass type used and is flat and non-polarization dependent over a long range (here: crown glass for visible emission).
Gratings have poor efficiencies due to the fact that the light is dispersed into various orders. Also, the efficiency is good for the specific wavelength for which the grid is designed, but falls off rapidly on both sides. Finally, the performance is alwa

An alternative dispersive element to the prism is a grating. Both transmission and reflection gratings are in use. Upon illumination with an incident beam, the periodic structure of the grating will cause the light to be deflected into various directions (by interference processes).

The straight direction (0th order) does not show any color dispersion. The 1st order is usually the direction of choice. Here, the light is spread into a spectrum, very similar to the dispersion in a prism. Nevertheless, there are more orders, e.g. 2nd and higher orders, but as well orders mirroring the 1st … nth order on the other side of grating normal. The art of grating fabrication is to concentrate as much energy as possible in one single order.

The grating also performs very differently for light that is polarized parallel or perpendicular to the direction of the lines (grooves). Whereas perpendicular waves can perform a rather efficient spectrum in the best case (depending on various parameters), the parallel wave drops significantly within two octaves away from the blaze wavelength – to about zero. As fluorescence is non-polarized, the total efficiency is described by the average between perpendicular and parallel efficiency. A drop to 30 % is common within 200 nm in the visible range.

This renders gratings as very inefficient and inappropriate dispersion devices for instruments where the efficiency (photon collection performance) is a critical issue – which is very much the case in confocal fluorescence microscopy. If implemented, a series of additional design elements are often appended that try to guide the lost photons to the sensors (sometimes referred to as "photon recyclers").

As a further complication, gratings show much more losses by stray light as compared to prisms.

Gratings have poor efficiencies due to the fact that the light is dispersed into various orders. Also, the efficiency is good for the specific wavelength for which the grid is designed, but falls off rapidly on both sides. Finally, the performance is alwa

Fig. 6: A grating produces a number of orders that contain a spectrum of the incident beam. The absolute efficiency is calculated by the ratio of the used order versus the incident beam.

Comparison of dispersers

As described in the text above, the various concepts of dispersion of fluorescence emission in confocal microscopes have advantages and disadvantages. A comparison overview is shown in the table below. Obviously, the prism is the best choice for the task.

Prism Grating Dichroics
Transmission s 0.95 0.75: 0.9 (blaze) – 0.6 (red) 0.9 (spec. dep.)
Transmission p 0.95 0.6: 0.8 (blaze) – 0.4 (red) 0.9 (spec. dep.)
Transmission (λ) white blaze depending specification depending
Higher order losses no yes no
Full spectrum for detection yes yes no (edges not infinite steep)
Independent of settings yes yes no (dispersion depends on actual elements)
Mechanically stable yes yes no (wheel, sliders ...)
Stray-light issue low high low

Comments

Help us to improve the Leica Science Lab and participate in our survey! Start Survey

Arrow Down

Arrow Up