1. Introduction

Traditionally confocal microscopes have used gas lasers as their light source. In 2000, Gerkin, et al. [1] was one of the first to report using a diode laser light source (406nm) for fluorescence confocal imaging. Since then, confocal microscope manufacturers have replaced some of their gas lasers with diode lasers [2], which offer several advantages. Diode lasers are physically more compact than gas lasers, making the microscopes smaller. The diode emission power – which is determined by the diode input current and operating temperature, both of which can be controlled extremely precisely by the operator – will be more stable than that of a conventional gas laser [3], the power output of which is determined by statistical processes. Thus, the more stable diode laser output will result in more reliable image information. Diode lasers are also less expensive and have a longer lifetime than gas lasers, resulting in fewer needed replacements. In addition, since reflected laser light is a major source of noise in confocal fluorescence images, reducing the reflected light intensity will improve the signal-to-noise ratio in the images. Diode lasers offer the advantage over gas lasers that by controlling the diode laser operating parameters – current and temperature – it is possible to alter the emission wavelengths and decrease the linewidth, thereby allowing more of the reflected laser signal to be filtered from the fluorescence image, resulting in improved image quality.

In confocal fluorescence imaging, a sample is first stained with a fluorochrome that binds selectively to the biological parts of interest [4], then the sample is placed in the microscope where a laser excites the fluorochrome, making it fluoresce. A barrier filter or dichroic mirror within the confocal optics theoretically removes all the laser light reflected from the sample, leaving only the fluorescence to create an image [5]. In practice, the filter transmits some of the reflected laser light, depending on the filter cut-off wavelength.

For example, a HeNe laser excites the fluorochrome Cy5 in a very narrow range around 633nm (see Fig. 1(a)). The Cy5 then fluoresces over the wavelength range 610–800nm (see Fig. 1(b)). However, this contains the reflected laser wavelength range. Since the fluorescence signal contains all the pertinent image information, the whole reflected laser signal must be blocked or, at the very least, minimized with a barrier filter or dichroic mirror. To do so, a 660nm barrier filter (18% transmission at 660nm) could be used, but this only passes the longest fluorescence wavelengths, eliminating information from a significant portion of the fluorescence signal (see Fig. 1(b)). Therefore, a slightly shorter wavelength (640nm) cutoff-filter was used for imaging with Cy5, but it transmits substantial portions of the reflected laser light (see Fig. 1(c)), resulting in increased spatial noise in confocal images, which in turn produces less image detail. Note that the unusually broad laser peak width shown in Fig. 1(c) results from displaying only the weakest 0.1% laser intensity – corresponding to that part of the reflected laser spectrum comparable to the fluorescence intensity. Thus, trade-offs exist with current filters and gas lasers – and even the currently used diode lasers (637nm, for dyes like Cy5).

To improve the image detail, the reflected laser spectrum collected by the objective lens must be separated from the fluorescence spectrum. Currently, multiple filter/dichroic mirror combinations (three or more) are used to accomplish this. However, the more optical elements in the light path, the more complicated the alignment procedure and the greater the loss in laser and fluorescence intensity. To compensate for this, higher power lasers and sensitive single photon counters are used, all of which increase the instrument cost.

An alternative way of separating the reflected laser and fluorescence signals without compromising either the incident laser or fluorescent intensities is to use a slightly shorter wavelength laser than either the HeNe or 637nm diode lasers currently available in commercial confocal microscopes. If the excitation wavelength is short enough (~600–610nm for Cy5), a single, short wavelength, long-pass filter can block the reflected excitation, increasing the fluorescence wavelength range detected, while reducing the noise due to the reflected laser light. Optical parametric oscillator (OPO) systems are capable of such output, having continuous wavelength tunability over hundreds of nanometers. However, the cost and space requirements of OPO-systems (which typically use Nd-YAG lasers) are similar to the large gas/solid state lasers that are standard with most confocal microscopes. Somewhat less expensive alternatives are external (or extended) cavity diode lasers (ECDLs), which, while offering greater output stability, have more limited wavelength tunability and other inherent limitations that make them less desirable for confocal microscopy. For example, changing the wavelength in a standard Littrow configuration ECDL requires modifying the diffraction grating angle, which causes the output beam angle to change. To compensate, a complex set of optics can be added [6], however, the more optics between the laser source and the sample, the lower the excitation intensity, the more delicate the alignment procedure, and the greater the cost. In Littman-Metcalf configuration ECDLs, the beam angle does not change with wavelength, but the output power is reduced considerably from the Littrow configuration. Both types of ECDLs are limited to approximately a 10nm tuning range for a single laser diode, which provides insufficient alteration in the excitation wavelength to eliminate overlap between the reflected excitation and the Cy5 fluorescence signals.

 

Fig. 1. (a) HeNe emission and Cy5 absorption spectra. The Cy5 absorbs the laser excitation with ~50% efficiency. (b) Cy5 fluorescence and 660nm barrier filter spectra show that much of the short fluorescence wavelengths are blocked by the filter, resulting in very faint fluorescence images. (c) The HeNe laser emission and 640nm barrier filter spectra show that the filter transmits a significant portion of the reflected HeNe spectrum. Only the bottom 0.1% intensity of the laser spectrum is shown, to approximately match the intensity of the Cy5 fluorescence, which was normalized to 1.0.

Download Full Size | PDF

In this paper, we describe a third alternative that, in comparison to other tunable laser systems, is very inexpensive and space efficient, while producing a sufficiently wide tunable wavelength range (~25nm) to reduce the reflected laser noise in the confocal images by about a factor of five. By capitalizing on the basic semiconductor physics property that lowering the temperature of a semiconductor increases its energy bandgap – i.e., shorter wavelengths – we shifted the semiconductor laser emission from 635nm to 609nm, while still efficiently exciting the Cy5 fluorochrome. This change in wavelength was sufficient for a single 640nm barrier filter to reduce the reflected laser intensity, while transmitting a greater range of the fluorescence spectrum and increasing its intensity. By so efficiently blocking the reflected excitation signal, the major contribution to the noise per pixel was reduced and the fluorescence signal-to-noise ratio was increased by slightly more than a factor of five. This increase in signal-to-noise was seen qualitatively in fluorescence confocal microscope images.

2. Materials and methods

A low power 3mW semiconductor, multiple quantum well (or superlattice), diode laser (Hitachi, 635nm – Thorlabs # HL6314MG) was used in these experiments. A LabVIEW program controlled the diode cooling, simplifying the difficulties normally associated with liquid nitrogen cooling. Since liquid nitrogen is frequently used in confocal microscopy sample preparation, it should be readily available in most confocal microscope facilities. Thus, only minimal cost increases are incurred due to cooling the diodes with the method described in this paper. To cool the laser diodes, the bare semiconductor laser chip was submerged in at least 99.998% pure dry nitrogen gas, the temperature of which was varied from room to liquid nitrogen temperatures. The gas temperature was controlled to within ±0.5°C and a T-type thermocouple, connected directly to the laser diode anode, monitored the diode temperature to within ±0.05°C. The diode laser output was measured via a fiber optic cable with a Coherent Laser Check power meter having a ±0.01µW resolution and an unattenuated 10mW saturation level. One end of the fiber optic was held in an aluminum block approximately 25µm from the semiconductor diode laser chip, also housed in the aluminum block. This allowed for reproducible fiber placement over the laser. A Perkin-Elmer 552 UltraViolet-Visible Spectrophotometer was used to measure the spectral response of the laser beam. Confocal fluorescence imaging with the cooled laser diode was performed with a Bio-Rad MRC 600 laser scanning confocal microscope mounted on an inverted IX-70 Olympus microscope. The diode laser beam was transmitted via a 200µm fiber optic cable to a custom collimator consisting of two aspheric lenses – replacing the custom mounted 30mW HeNe laser and its downstream optics – and ported into the Bio-Rad optics.

 

Fig. 2. Diode current vs. diode voltage. The solid black curves are constant power, with the 60mW curve the upper bound on power applied to the diode.

Download Full Size | PDF

3. Experimental results

3.1 Diode current-voltage dependence on temperature

To avoid fracturing the diode’s superlattice from thermal gradients within the superlattice, it was necessary to apply no more than a maximum power of 60mW, determined at approximately room temperature. As a semiconductor is cooled, the number of thermally generated free charge carriers decreases, resulting in a higher voltage across the diode, which, coupled with the power limit, requires the applied current to be reduced as the diode is cooled. To ascertain how this decreased current would affect the diode’s output light power as the laser was cooled, we extracted – at each temperature – the effective reverse-bias saturation current (${\mathrm{I}}_{\mathit{\text{saturation}}}^{\prime }$ ) and ideality factor (n) from the diode equation [7] fit to the diode’s current-voltage data (see Fig. 2):

$I_{\mathit{diode}}=I_{\mathit{saturation}}^{\prime }\left({\text{e}}^{q{V}_{\mathit{diode}}⁄n{k}_{B}T}-1\right)$

where I _diode is the measured diode current at a specific applied diode voltage (V _diode ); q is the elementary electron charge; k_B is Boltzmann’s constant; and T is the diode operating temperature (in Kelvin). In laser diodes, where significant radiative recombination of charge carriers occurs, the effective saturation current depends inversely on the effective lifetime of the charge carriers [8], i.e. the effective time between collisions. If more radiative recombinations occur at lower temperatures where there are fewer charge carriers, the effective carrier lifetime must decrease, resulting in an increase in the effective saturation current. Additionally, when more recombinations occur, the ideality factor will also increase and become much greater than one, because the ideality factor reflects the diode’s departure from ideal behavior (which assumes no radiative recombinations). From our data, the saturation current increased by several orders of magnitude with decreasing temperature, while the ideality factor increased by more than an order of magnitude (from 5 to 100) over the same temperature range. These results indicate that as the diode operating temperature is lowered, the number of radiative recombinations increased. This in turn indicates that the diode’s output photonic power increased as the diode was cooled, offsetting the decrease in applied current due to the constant power requirement.

3.2 Laser power output dependence on temperature

To quantify the photonic power growth with decreasing temperature, a laser power meter was used to measure the output diode power as a function of temperature (see Fig. 3). All power measurements were performed with a 61cm flexible fiber optic light guide with a 52.4% transmission in the 600nm range. The measured powers in Fig. 3 were adjusted for transmission losses in the fiber optic and correspond to the actual power at the diode before entering the fiber optic. The measured power increases were so large at the two highest currents that the laser power meter saturated at its rated 10mW with the diode laser cooled to between -30°C and -40°C. The physical arrangement of our measurement apparatus did not permit the use of an attenuator and thus no quantitative power data was possible below -30°C for the 24mA and 27mA curves, though a quadratic curve fit provides a rough estimate of the output power behavior in this region. Generally, as the diode was cooled, its laser output power was greater than at room temperature. This turned out to be of paramount importance in fluorescence confocal imaging, because as the diode was cooled its emitted wavelengths decreased (see section 3.3), which excited the fluorochrome slightly less efficiently, but the power increase compensated for the decreased fluorochrome excitation (see the confocal images of Section 3.6).

 

Fig. 3. Laser power as a function of temperature, showing a power maximum around -80°C and parabolic behavior down to -120°C. The power meter saturated at output powers above ~20mW (adjusted for fiber optic transmission losses).

Download Full Size | PDF

3.3 Spectral response (wavelength shift)

Upon cooling the laser diodes, the emission color changed markedly from red to orange. Since the laser wavelength is directly related to the semiconductor bandgap energy, which increases (shorter wavelength) with decreasing temperature, the laser should emit at shorter wavelengths upon decreasing its operating temperature. This was observed with a spectrometer (see Fig. 4(a) for three representative spectra) and the peak wavelength was found to vary linearly with temperature (Fig. 4(b)). The advantage of this linearity is that it allows precise temperature-controlled wavelength tunability over a ~25nm range and a ~220°C temperature range (room temperature to liquid nitrogen). Though this data was taken with the diode current at 20mA, the wavelength shifts were identical for all other currents.

 

Fig. 4. (a) Spectral response of the laser diode at 20mA and three different temperatures, showing the shift to shorter wavelengths with lower diode operating temperature. Intensity is in arbitrary units. (b) Peak wavelength versus diode operating temperature is linear, demonstrating the wavelength tunability of cooled diode lasers.

Download Full Size | PDF

In confocal microscopy, this temperature-dependent wavelength shift causes the laser emission to move away from the barrier filter edge for sub-ambient diode operation (see Fig. 5(a)), resulting in less reflected laser light in the fluorescence image. This reduces the spatial noise level in the confocal images. The shift to shorter laser wavelengths excites the fluorochrome in a different portion of its absorption spectrum (see Fig. 5(b)), however, the increased laser power at the lower diode operating temperature still efficiently excited the fluorochrome, producing a more intense fluorescence signal (see Fig. 6(a)). To compare the decrease in noise level when operating the diode laser at liquid nitrogen versus room temperatures, the reflected laser signals passed by the barrier filter were calculated for the same fluorescence signal levels. Figure 6(b) shows that, for equal fluorescence levels, the relative noise level dropped by more than a factor of five when the diode laser was cooled to liquid nitrogen temperature.

 

Fig. 5. (a) Diode laser emission and barrier filter spectra, showing the barrier filter almost completely blocks the emission from a diode cooled to -196°C, while transmitting more when the diode operates at room temperature. (b) Laser diode emission and Cy5 absorption spectra, showing that at -196°C the diode laser excites the fluorochrome nearly as efficiently as at 23°C.

Download Full Size | PDF

 

Fig. 6. (a) Fluorescence signals produced by exciting Cy5 with the laser diode operated at constant current, but two different temperatures: 23°C and liquid nitrogen temperature. (b) Relative noise level – for equal fluorescence signal maxima – due to reflected laser intensities after passing through the 640nm barrier filter with the diode operated at 23°C and liquid nitrogen temperature. The noise (reflected laser light) is decreased by a factor of 5 at -196°C, resulting in a factor of 5 improvement in signal-to-noise ratio.

Download Full Size | PDF

 

Fig. 7. (a) Laser power output as a function of diode current for various temperatures. The threshold current is the x-axis intercept of each line. The threshold current decreased by 75% as the temperature was decreased from room temperature to -170°C. (b) Threshold current as a function of temperature, showing the usual exponential temperature dependence.

Download Full Size | PDF

Thus, by adjusting the laser’s operating temperature, its emission wavelength can be tuned to excite fluorochromes having excitations within ±40nm of 640nm, such as YO-PRO-3 (613nm), Syto 17 (621nm), Bodipy 630 (625nm), Alexa 633 (632nm), TO-PRO-3 (644nm), Bodipy 650/665-X (646nm), Cy5^™ (649nm), Alexa 647 (650nm), APC (651nm), Alexa 660 (664nm), Cy5.5^™ (675nm) and Alexa 680 (679nm) [9]. Such spectral tuning would eliminate the costly need for many diode lasers emitting at different wavelengths and allow the laser output to be tuned to eliminate the reflected laser light with a barrier filter or dichroic mirror.

3.4 Threshold current

In addition to the increase in diode output power upon cooling, the current needed to produce lasing dropped significantly, indicating an increase in laser operating efficiency. Since it is difficult to visually or electronically discern when lasing first begins, the threshold current is defined as the x-intercept of the laser output power versus diode current for a specific operating temperature, as shown in Fig. 7(a). The threshold current decreased 75% over approximately 150°C as the diode laser was cooled from room temperature. Plotting the threshold currents, as a function of temperature, produced the standard exponential temperature dependence (see Fig. 7(b)). The drop in threshold current with decreasing temperature indicates that less input power may be needed to produce lasing, however, to determine this correctly, the product of threshold current and corresponding input voltage must be calculated (see next section).

3.5 Efficiency versus temperature

The diode efficiency is defined as the ratio of the laser’s measured output power to its input power (=applied current times applied voltage). Cooling the laser increased the diode efficiency to a remarkable 42% (see Fig. 8), four times greater than its rated efficiency (~10%) at room temperature. The efficiency had a maximum around -80°C to -90°C. Thus, not only did the output power increase, but the conversion of electrical power (or input power) into photonic output power also became more than four times as efficient.

It is important to note that the efficiencies obtained close to room temperature at the rated 20mA operating current (see Fig. 8) are greater than 10%, because the diodes were force-cooled to 21.8°C, the highest temperature attainable when dry nitrogen was flowing through our experimental setup. Without such cooling, under normal circumstances, resistive power losses would internally heat the semiconductor chip and drive the laser’s operating temperature above ambient, lowering the laser power output and efficiency to the manufacturer’s rated 10% maximum. In confocal microscopy, such operating temperature changes would cause uncontrolled shifts to slightly longer emission wavelengths, increasing the amount of reflected laser power passed by the barrier filter, reducing the signal-to-noise ration in the confocal images.

 

Fig. 8. Diode efficiency as a function of temperature at various operating currents. At around – 80°C, the diode efficiency was the greatest, independent of operating currents. The precise maximum efficiencies for the 24mA and 27mA data sets are not known, since the power meter saturated above 20mW (adjusted for fiber optic transmission losses).

Download Full Size | PDF

3.6 Confocal microscopy

Since cooling the diodes led to improvements in their operation, we attached the laser diode – via fiber optic cable – to a Bio-Rad MRC 600 confocal microscope and took a series of fluorescence images with the diode cooled to several different temperatures. Figure 9 shows the projections of z-series scans created from fluorescence images of Cy5-soaked filter paper as the diode operating temperature decreased from 6°C (somewhat close to room temperature), to -49°C and -97°C (near the maximum in diode output power), to -196°C (the greatest separation between the fluorescence and reflected laser wavelengths).

 

Fig. 9. Fluorescence confocal microscope images of filter paper fibers soaked in Cy5. The images were taken while operating the diode laser at different subambient temperatures ranging from 6°C (top left), to -49°C (top right), to -97°C (bottom right), to -196°C (bottom left), which shows the greatest contrast.

Download Full Size | PDF

The fluorescence produced by the diode at the two middle temperatures was so intense, that the photomultiplier detector was saturated over much of the image. With current confocal microscopes, this would be less of a problem, as the photomultiplier dynamic ranges are greater than the one used here, from the early 1990s. Since the signal-to-noise ratios increased upon decreasing the diode operating temperature (Fig. 6(b)), more detail should be apparent in each image created by successively colder operation of the laser diode. Qualitatively comparing the four photographs, the -196°C image contains more detail than the 6°C image. At -196°C, the diode laser wavelengths were just at the filter cut-off wavelength (see Fig. 5(a)), resulting in the least amount of reflected laser light of the four images. In comparison, at 6°C, a significant portion of the reflected laser light passed through the filter (see the 23°C curve in Fig. 5(a)), washing out the fluorescence detail. Thus, of the four photographs in Fig. 9, the -196°C image contains the most fluorescence information … even though the images at -49°C and -97°C appear the brightest, they contain more reflected laser light and hence less fluorescence information.

4. Conclusion

A major problem with fluorescence confocal microscope images is distinguishing between the fluorescence signal and the reflected laser signal, especially with older microscopes, where the He-Ne sources operate in the tens of milliwatts range, allowing significant reflected laser intensity to be transmitted through the long pass barrier filter or dichroic mirror and be observed in the image. One possible way of solving this problem is to use a long pass filter that blocks most of the reflected laser signal, but that resulted in eliminating a significant portion of the fluorescence signal. A potentially more informative technique described in this paper is to cool the laser diode and thereby shift its emission wavelengths to shorter than the barrier filter cutoff, while still efficiently exciting the fluorochrome.

We showed that the laser output power of a cooled 635nm semiconductor diode laser was approximately a quadratic function of temperature, with the maximum output power about five times greater than at room temperature. At the same time, the threshold current decreased with decreasing temperature, while the device efficiency increased by a factor of four over the diode’s room temperature operation. Most significantly for confocal microscopy, the laser emission wavelength had a very linear temperature dependence, allowing precise wavelength tunability within more than a 25nm bandwidth. This ability to shift the emission wavelength to the barrier filter cutoff had the advantage of increasing the signal-to-noise ratio by more than a factor of five and produced an image with significantly more fluorescence and less reflected laser light, resulting in an image qualitatively containing greater contrast and detail. The disadvantage of the wavelength shift is that the fluorochrome (Cy5 in our experiments) was excited slightly less efficiently than at room temperature. However, the increase in laser output power upon cooling compensated for the decreased excitation so much that the fluorescence signal saturated the detector at the greatest diode output powers.

While the greater contrast and detail in our experiments is useful, current state-of-the-art confocal microscopy uses multiple dyes to obtain even more information from a single sample. In this situation, several different lasers are used to excite the various dyes, while multiple filters and dichroic mirrors isolate the different fluorescence signals, compounding the problems of distinguishing between the reflected laser and fluorescence signals. Tuning the emission wavelengths of cooled diode lasers to the various filter and dichroic mirror cutoff values would minimize the laser contamination in the fluorescence images, improving the signal-to-noise ratio, thereby increasing the information contained in the images.