Surprises of Medium-Wave IR LEDs Based on А3В5Heterostructures
ecently, the tendency of overcoming of "hydrophobia" caused by the misunderstanding of the specific peculiarities of medium-wave IR light emitting diodes has been outlined and increasingly more companies have declared about the commencement of production of "light emitting" gas analyzers (GA). In the meantime, for the certain part of instrument developers such LEDs are still some "exotics" which requires additional explanations. In the offered article we will try to give such explanations and consider the new properties of medium-wave LEDs which have been recently discovered.
Usually, the developers who study the specifications received from LED manufacturers for the first time are surprised because of the unordinary high values of the half-width of electroluminescence (EL) spectrums which are equal to 0.3–0.7 µm depending on the wavelength. For the optical non-dispersion (NDIR) GA such broad radiation spectrum causes problems due to the cross sensitivity, because the simultaneous absorption of several gases influences on the value of the light flux which passed through the analyzed mixture. For example, when measuring the concentration of carbon monoxide (analytical wavelength is 4.7 µm) using the measurement of radiation attenuation from the light emitting diode with the maximum wavelength of 4.7 µm during its transmission through the gas, the absorption at the wavelength of 4.3µm turns out to be very significant; it is caused by the presence of atmospheric carbon dioxide (concentration 0.03% by volume) in the optical path . It is commonly supposed that the LED spectrum width is equal to 1.8 kT (k is the Boltzmann’s constant, T is the temperature), and therefore the task of spectrum narrowing can be solved at the expense of LED cooling. However, very likely that the use of coolers will not bring optimism to the developers and end users because during the use of cooling systems the main advantages of LEDs – their low energy consumption and overall dimensions – are leveled. In order to eliminate the cross impact of СО and СО2 on the GA parameters the "spectral" tricks are used; such as, for example, application of narrow-band interference filters (usually, ∆λ=50–100 nm), Fabry-Perot resonators, the mode structure of which repeats the spectrum of one of the gases , or filters with "negative filtration", which consist of the cuvette with one of the gases with high concentration . Of course, the listed methods are applicable only for single-component GAs; the use of diffraction grating is the most convenient method for multi-component analysis under the conditions of GA immovable parts. In this case, large width of EL spectrum is the positive property which allows creating the great amount of spectral measurement channels .
The next surprise can be the fact that according to the publications of some authors the conversion coefficient or, in other words, amount of the light energy released in unit time considerably exceeds the conversion coefficient in the pulse mode for the continuous or quasi-continuous (in other words, with the porosity 2) operation. These authors explain the discrepancy of the values of continuous and pulse EL power reaching several times with the same current value by the heating of LED active region during the current passage. Indeed, LED with non-optimal design (as a rule, these are LEDs with the point top contact and p-n junction which is distant from the holder) are subject to the intense heating of p-n junction during operation which in reality causes the decrease of conversion coefficient . However, the intense LED heating also causes the noticeable variation of the width of forbidden band and, of course, EL maximum wavelength. Nevertheless, such spectrum variations or differences in pulse and continuous modes were not registered by these authors.
In other types of LEDs with good heat pick up and radiation coupling through the transparent substrate, the EL spectrums (λmax=3.3 µm) did not depend on the current within the limits from 0.01 to 500 mA , which corresponded to the conceptions of insignificant Joule heating of such LEDs and identity of continuous and pulse power, at least, in the region of low currents. In order to estimate the temperature influence on the position of maximum LED radiation, we can use the specifications for LED43Sr radiating at the wavelength of 4.2µm at 300 K as example . Increase of the temperature (more precisely, ambient temperature) from 25 to 90°C corresponds to the double decrease of power of such LED and the EL maximum wavelength grows from 4.15 to 4.4 µm. It follows that it is impossible not to notice such significant variation of EL spectrum maximum during the experiment. The simple conclusion from the consideration of aforementioned examples consists in the fact that when selecting the components for GA it is necessary to study attentively and analyze all parameters declared by the manufacturer including the temperature dependencies of LED spectrums and power in continuous and pulse modes. Of course, the LEDs, which have the data given with the concordance of pulse and continuous EL power, at least in the region of low currents, have higher credibility.
For high-precision GAs even small (and often unavoidable) temperature variations of EL wavelength give rise to the serious concerns of developers . In the meantime, in the immersion LED with flip-chip construction such variation does not have hysteresis and is reproduced upon the multiple thermocycling, therefore, it can be taken into account in the analytical expression for GA transfer function and the high precision of measurements can be reached. Such approach made it possible to reach the detection limit of 50–100 ppm with the volume of analyzed gas (CO2, C2H4 or СО) less than 10 ml and averaging time of several seconds which meet the requirements of the majority of tasks for portable GAs .
The term "negative" in relation to the physical phenomenon does not always mean something useless or unnecessary for the technical applications. In this context, the negative luminescence (NL) only means that the main flux of (non-equilibrium) photons is directed not towards the observer but from him. Such situation was detected for the first time during the research of recombination radiation in indium antimonide and afterwards it was reproduced in many diode structures based on CdHgTe, InAsSb, InAs and more complex materials – supergratings having low leakage currents . For NL initiation it is sufficient to displace LED in backward direction and LED will start absorbing photons more than radiating them outwards within the limits of its absorption spectrum. The distribution of intensity of luminescent radiation is shown in figure as the example upon the simultaneous connection of all elements of monolithic array 3х3 with the active layer of InAs (λmax=3.4 µm, 300 K) to the external power source . In this experiment the diode elements were united into two groups consisting of 4 and 5 elements respectively as it is shown in the insertions to figure. IR image of array surface obtained at the direct current ≈- 25 µA in each of 4 non-diagonal elements, which operate under the NL conditions or conditions of decreased radiation capacity (dark squares) in comparison with the equilibrium background, and at the current ≈20 µA in each of diagonal and central elements, which operate under the NL conditions (light squares, increased radiation brightness in comparison with the background) is shown in the right insertion to figure. In other experiment (see left insertion) four diagonal and central array elements operated under the NL conditions (Ipixel≈-20 µA), and four remaining elements – under the EL conditions (Ipixel ≈ 25 µA). Amplitude of the total current in both cases of the formation of IR image of the type "board for noughts and crosses’ was 100 µA.
From the data in figure it follows that at the currents close to 100 µA the intensities of NL and EL practically concurred and at high currents the small dependence of NL intensity on the current occurred – it is well known peculiarity of NL devices which is caused by almost complete extraction of carriers from the LED active region [9, 10].
Besides NL use, for the creation of "cold" screens in cooled photodetecting systems based on the arrays with large size, for the testing of such arrays [8, 9] and obtainment of combined, "temperature-independent" optical signal , NL is important for the studies and forecasting of the parameters of photodiodes (PD), which have one of the contacts on the illuminated semiconductor surface with the high value of bulk resistance. Thus, for example, the studies of the spatial distribution of NL in PD based on InAsSb revealed the photocurrent concentration near the contact and disclosed the main reason of their low current photosensitivity – incomplete collection of photogenerated carriers . These studies allowed viewing the construction of medium-wave PD in a new way  and creating the high-efficiency optical gas sensors which can operate even at high temperatures .
The distinctive feature of medium-wave LEDs is sublinearity of their Watt-Ampere characteristics even at insignificant values of currents. Decrease of the efficiency with the current increase is caused not only by the aforementioned LED heating during its operation but by more fundamental reasons as well – so-called Auger recombination which has such name after the name of the discoverer of basic mechanisms of radiationless recombination of non-equilibrium carriers in the narrow-bandgap compounds of InAs type. The speed of this recombination is proportional to the cube of the concentration of injected carriers, and therefore the developers and manufacturers of medium-wave LEDs prefer using quite thick semiconductor layers, in which the charge carriers are "spread" throughout the active region. As a result, relatively low concentration of injected carriers and low speed of Auger recombination are reached accordingly. Of course, the upper bound is selected for the thickness of active region on the basis of the capability to couple radiation out of this region taking into account the radiation self-absorption. As a result, the optimal thickness of LED AR (active region) often turns out to be comparable to the optimal thickness of AR in PD. In other words, the efficient medium-wave LED often simultaneously turns out to be efficient PD. This circumstance was taken into account in the selection of the common design for many medium-wave LEDs and PDs with broad reflecting anode [4–8]. As a result, the expenses connected with production were significantly decreased due to the fact that the same post-growth operations and accessories are applicable for PDs and LEDs including, for example, templates for photolithography. The described "duality" of the properties of medium-wave diodes can turn out to be useful for the projecting of GA as well. Thus, for example, in  during the photometric measurements including the gas analysis, it was suggested to use the same diode in the capacity of detector and then in the capacity of radiation source by turns. Such "ruse" will be especially useful for the analyzers with large amount of optically connected LEDs and PDs  because it significantly increases the number of useful signals which can be used for the self-calibration of GAs with the relevant mathematical processing of digitized signals – this task is still topical.
But probably the most pleasant surprise of medium-wave LEDs is their capability to operate in the capacity of thermal pump with low displacements of p-n junction. Graduate students and workers of Massachusetts Institute of Technology (USA) established that the LEDs manufactured by the company "IoffeLED" based on p-n structures InGaAsSb with the near wavelength of 2µm have the capability to take part of the thermal energy from the crystal lattice and transform it into the energy of photons . As a result of interaction of phonon and photon fields, the LED efficiency, which is determined as the ratio of the energy released from LED in the form of photons to the expended electric energy, turns out to be considerably higher than one (230%, 135ºС). Unfortunately, at the temperatures, which are close to the room temperatures, the "super high" coefficient of efficiency occurs only in the region of low currents which are less than several picoamperes . In order to implement the measurement device using such low LED supply currents, the PDs with extremely low intrinsic noises will be required, for example, PDs cooled to the temperatures close to 77 K . It compels us to consider the application of the effect of thermal pump for practical purposes, for example, in inexpensive portable "light emitting" GAs, untimely, as for now. So we can just to rely on the diligence and commitment of the graduate students who are inspired by the beauty of the new physical phenomenon and ready to "mine" and then defend and put into practice the new scientific provisions in the major 01.04.10 – Physics of Semiconductors.
Author expresses gratitude to the workers of the group of diode optical couplers of the Laboratory of Infrared Optoelectronics of the Ioffe Physical-Technical Institute of the Russian Academy of Sciences for the useful recommendations during the article preparation.