All mid-infrared detection products of NLIR are based on sum-frequency generation (SFG), which is a highly sophisticated nonlinear optical process where two-photons annihilate in the creation of a new photon with the same energy as the sum of the originals.
Many know the special case of SFG called second-harmonic generation, which is used most widely in laser pointers to generate green light from a 1064 nm laser diode. In our products, we use SFG differently in a process called upconversion where focus is on changing the wavelength of incoming mid-infrared (MIR) light to near-visible light.
Nonlinear optical processes are in general driven by high-intensity electrical fields inside media that exhibit large nonlinear coefficients and small losses, for example LiNbO3 crystals. For upconversion, in which three electrical fields participate, one of these fields must be of sufficiently high intensity. Upconversion has been known for many years, but the requirement of a high-intensity laser field has made the technology either too expensive or too inefficient to commercialize.
The schematic pointed to by the arrow illustrates energy conservation in the upconversion process where one photon from the high-power laser field and one MIR photon combine and become one near-visible light photon; the length of the arrows signify the energy and, hence, wavelength. The reason for the term “Virtual state” in the top of the diagram is that the photons are not absorbed to a physical state in the crystal (as is the case for electronic absorption) but only interact with bound electrons through electric forces.
After upconversion, the near-visible light can be detected with a detector suitable for the application in mind, for example a CMOS array in a grating spectrometer (like the 2.0 – 5.0 µm Spectrometer in “Products”), or a very fast or sensitive point detector (like the Single-wavelength Detectors).
The products of NLIR have exceptional noise properties. Many sources of noise known from conventional MIR detectors are avoided but a new one is introduced during upconversion.
The requirements for upconversion, which limit the achievable efficiency and bandwidth, inherently filter out thermal noise from the surroundings. Only noise in the detectable bandwidth that co-propagates with the signal is upconverted and makes it to detection. The rest of the thermal noise from the surroundings that would be detected by a conventional MIR detector is simply not upconverted. Similarly, a conventional MIR detector also has a significant amount of internal noise because of its own temperature, which is the reason why many MIR detectors are cryo-cooled. This noise contribution is also largely avoided after upconversion because hardly any thermally emitted photons exists at the upconverted wavelengths at room temperature.
The upconversion process itself is perfectly noiseless. It has been showed in the literature that even on a quantum mechanical level, an upconverted photon is an exact copy of the longer-wavelength original except for the wavelength. This means that the upconversion process itself does not contribute with any noise to the detection scheme.
However, two parasitic processes take place inside the crystal during upconversion. Firstly, the crystal itself has a temperature that causes it to emit photons randomly; those that are emitted in the same direction as the signal are upconverted and added to the detection as noise. This noise dominates above 3.0 µm and grows with wavelength. Secondly, photons from the high-power laser decay through spontaneous downconversion (yet another nonlinear process) to end up in the same direction as the signal, and when upconverted they are detected as noise. This latter process is referred to in the literature as upconverted spontaneous parametric downconversion (USPDC) noise, and it dominates at wavelengths below 3.0 µm.
The two new sources of noise introduced in the upconversion scheme are significant enough to be observed in the 2.0 – 5.0 µm Spectrometers at sampling frequencies < 0.1 Hz. In the Single-Wavelength Detectors, the contribution varies depending on wavelength and bandwidth, but its effect is always included in the specifications.
An important part of the physics behind upconversion is the time constant of the nonlinear interaction. A common misunderstanding among non-experts is that the conversion process happens at the speed of light. While it is true that the light is not slowed down by other than a factor of the linear refractive index during the conversion process, the interaction itself has little to do with how fast light is moving. The time constant of the conversion process is determined by how fast the bound electrons in the material respond to the force of an applied electric field. Naturally, the electrons do not react infinitely fast, but their response is so rapid compared to all other dynamics involved that for (almost) any application, their response can be considered instantaneous. Hence, the upconversion process itself does not cause time lag, or jitter, or blurring of any kind.
However, it turns out (strictly speaking) that dispersion in the crystal in principle causes time blurring on very short time scales. In the most sensitive Single-Wavelength Detectors that NLIR manufactures, the effect will start becoming notable at modulation frequencies above 100 GHz. In the more broadband devices, it will be at even higher modulation frequencies. Consequently, none of the Single-Wavelength Detectors are affected by the dispersion-induced time blurring. The Spectrometers are only affected in terms of slightly reduced sensitivity for input pulses shorter than 100 fs; the measured spectrum is not affected.
The efficiency of the upconversion process is determined by the intensity of the high-power laser inside the nonlinear crystal and by the desired optical bandwidth. A higher power and a narrower bandwidth gives a higher efficiency. For a wide bandwidth of 2.0 – 5.0 µm, the conversion efficiency is approx. 0.001, which for many applications, together with an appropriate near-visible light detector, still outperforms a standard MIR detector alternative in terms of noise-equivalent power. If a narrower bandwidth is desired, for example 3.3 – 5.0 µm, the conversion efficiency is > 0.01, and for a bandwidth of only 50 nm around 3.0 µm the conversion efficiency is > 0.1.