NONLINEAR OPTICS: All-optical method upconverts 2-D images

CHRISTIAN PEDERSEN, JEPPE SEIDELIN DAM, and PETER TIDEMAND-LICHTENBERG

While low-cost, high-sensitivity, and high-speed image sensors (digital cameras) exist for the visible and near-IR (NIR) spectral range, imaging is far more difficult in other spectral regions of interest—for example, in the IR and terahertz ranges. We have developed a new and efficient approach for frequency conversion of images from one part of the electromagnetic spectrum into the visible or NIR spectral range. The approach is based on enhanced nonlinear upconversion of radiation from one part of the electromagnetic spectrum to the visible or NIR spectral range, allowing the use of existing state-of-the-art NIR cameras. The frequency-conversion process can have further built-in functionality like spatial or spectral filtering, and thus provide functional spectral imaging.

Improved upconversion

Upconversion has been investigated for decades, and was a particular focus in the late 1960s to mid-1980s.^1-3 One important parameter limiting the practical applications of the upconversion process has been the conversion efficiency. Efficiency is especially important for any application requiring high sensitivity or high speed. One approach to raise the nonlinear conversion efficiency is to exploit the high intracavity field of a laser resonator. This was first demonstrated in 1978, but the efficiency in that case was limited by intracavity loss originating from absorption in the nonlinear crystal.⁴ The upconversion efficiency was reported to be 0.38% at best.

Our recent results show that it is possible to convert an object field at one wavelength into a new spectral region with high conversion efficiency (40%), even for object fields containing spatial information.⁵ The efficiency increase was obtained by a combination of today’s high-finesse solid-state laser cavities and efficient periodically poled nonlinear crystals. A third important aspect of the recent results is the development of highly efficient cameras that allow for fast, low-noise detection. Our method is all-optical and performs upconversion of full two-dimensional (2-D) images. This is in contrast to more-conventional methods that rely on x-y scanning devices.

In future applications, we foresee a range of interesting possibilities in the mid-IR, IR, and terahertz regions. Much effort today is directed toward detection of gaseous components such as nitrogen oxides (NOx) and carbon dioxide (CO₂) in environmental pollution monitoring, in the optimization of combustion systems, and for catalytic reduction methods—for example, to reduce NOx emissions.⁶ These gaseous components typically have fundamental absorption bands in the 3 to 6 µm spectral range where 2-D imaging with high spectral resolution is difficult. Similarly, in the terahertz range, simple 2-D imaging will allow the user to see through clothes and packing materials and identify dangerous items or substances underneath. Terahertz imaging systems will also be able to detect contaminants and impurities in consumer food products—for example, glass splinters in chocolate bars or metallic fragments in tobacco.

Nonlinear imaging and filtering

In the basic upconversion scheme, an input beam comprising several spatial frequencies originates from a coherently illuminated object—in other words, not necessarily a Gaussian field distribution (see Fig. 1). The beam is Fourier-transformed to the Fourier plane using a lens (f). At the Fourier plane, the object beam interacts with an intense laser beam (with a Gaussian intensity distribution shown as red) inside a nonlinear medium, through a sum-frequency generation process (SFG). In the small-signal plane-wave approximation, a beam proportional to the product of the two interacting fields will emerge at the sum frequency. A second Fourier-transforming lens (f1) provides the upconverted image at the image plane.

A detailed analysis of the image-formation process shows that the upconverted image is indeed a magnified replica of the object beam, but convolved, or blurred, with a Gaussian spatial function. The blurring depends on the Gaussian beam radius (W₀) at the Fourier plane and the focal length of the Fourier-transforming lens f1. In the two extreme cases where W₀ is either much bigger or much smaller than the spatial features of the object beam at the Fourier plane, either undistorted image upconversion or TEM₀₀ filtering takes place, respectively.

Experimental results

The actual upconversion setup is realized as an intracavity design in which the SFG crystal—periodically poled potassium titanyl phosphate (pp-KTP)—is situated inside the cavity of a diode-pumped, high-finesse 1342 nm vanadate (Nd:YVO₄) laser, thereby exploiting the high circulating power of the Gaussian beam (see Fig. 2).

FIGURE 2. An experimental setup used for nonlinear image conversion and filtering is built around a pp-KTP nonlinear optical crystal.

When pumping with 2 W of power at 808 nm, the intracavity power at 1342 nm reaches roughly 120 W. This enhances the upconversion efficiency by two orders of magnitude compared to single-pass configurations. The object field is generated by coherently illuminating a transmission mask forming a cross with a collimated external-cavity diode laser (ECDL) at 765 nm (see Fig. 3). The wavelength of the generated upconverted image is 488 nm, which corresponds to the sum frequency of the two interacting beams at 1342 and 765 nm, respectively.

FIGURE 3. A transmission mask (a) is positioned at the object plane. A direct image of the mask coherently illuminated at 765 nm was taken (b). The theoretically calculated light distribution (based on dimensions from the direct image) at a 488 nm wavelength at the final image plane (c) can be compared to the measured 488 nm upconverted cross at the image plane (d). A reconfigured experiment shows the improved resolution when using a larger-diameter Gaussian beam 1342 nm (e). All images have been colored to reflect their visual appearance.

The power transmitted through the mask was measured to be 15 mW and the converted blue image contained 6 mW of power. Thus, 40% power-conversion efficiency or 25% quantum efficiency was obtained. To the best of our knowledge, this is the highest reported upconversion efficiency of a real 2-D image under a continuous-wave condition.

Because a power-conversion efficiency of 40% corresponds to an average decrease in the object field of 14%, we conclude that the underlying assumption of the small-signal approximation is valid. The Gaussian blurring of the upconverted image closely matches the theoretically calculated image.

Theory predicts an increased resolution if the radius of the Gaussian 1342 nm beam is increased. This is indeed the case: Moving the pp-KTP crystal to a position near the mirror, M4, where the beam diameter is significantly bigger, clearly improved the spatial resolution of the cross. However, increasing the diameter of the 1342 nm beam decreases the conversion efficiency. With all other parameters equal, optimization of one of the two parameters, resolution or efficiency, can be accomplished only at the expense of the other. One route to circumvent this limitation is to increase the circulating power of the intracavity beam or to use more-efficient nonlinear crystals.

A second prediction from the theory is low-pass spatial filtering, or beam cleanup. Our experiment begins with a poor-quality beam from an ECDL at 765 nm at the object plane (see Fig. 4). The spatial extent of the 1342 nm Gaussian intracavity beam at the Fourier plane is arranged so that the 1342 nm beam diameter is smaller than the spatial extent of the 765 nm object beam. In this case, TEM₀₀ filtering is expected, which indeed is the case. At the image plane, even with the central part of the 488 nm beam overexposed, no signs of secondary spatial features were seen, indicating a strong TEM₀₀ selection. This was further confirmed by actually measuring the M² values of the 765 nm beam and the upconverted 488 nm beam: The measured M² values were 2.4 and 1.05, respectively. The conversion and filtering efficiency of the process was measured to be 19%.⁷

FIGURE 4. The spatial distribution of a poor-quality tapered diode laser is shown at the object plane (left). Measured intensity distribution of the object beam (center, green) and the Gaussian intracavity beam (center, black) at the Fourier plane shows the smaller size of the Gaussian beam. The converted and spatially filtered upconverted beam at the image plane shows the resulting TEM00 beam.

An interesting feature is that the spatial filtering produces a true Gaussian shape and does not suffer from scattering, as would be encountered if a hard aperture (such as a pinhole) were used for spatial filtering. Note that because ECDLs now approach 10 W in output power when tapered diode-laser technology is used, scattering from a hard aperture could cause unwanted feedback to the laser unit.

Outlook

Modern technologies, including high-finesse cavities for enhancement, flexible and efficient nonlinear crystals, and CCD camera technology have given this field new possibilities and eliminated significant drawbacks that have limited former work.

Future work will be directed toward the mid-IR region or the terahertz range, where 2-D detection with high spectral resolution is more difficult or even impossible with existing technologies. In this context, it is worth mentioning that the upconversion process adds very little noise to the signal (the noise is primarily upconverted thermal radiation). In addition, extremely fast phenomena can be upconverted, because the conversion process depends on the nonlinear crystal’s electronic response—in other words, the near-instantaneous movement of electrons inside the crystal. These features make this technology suitable for the detection of very fast processes.

REFERENCES
1. R.A. Andrews, J. Quantum Electronics, p. 548 (1969).
2. A.H. Firester, J. Appl. Physics 41(2), p. 703 (1970).
3. W. Chiou, J. Appl. Physics 42(5), p. 1985 (1971).
4. J. Falk and Y.C. See, Appl. Phys. Lett. 32(2), p. 100 (1978).
5. C. Pedersen et al., Optics Express 17, p. 20885 (2009).
6. D.J. Stothard et al., Optics Express 12, p. 947 (2004).
7. E. Karamehmedovic et al., Appl. Phys. B 96, p. 409 (2009).

Christian Pedersen is the group leader of the Optical Sensor Technology group at DTU Fotonik, the department of photonics engineering at Technical University of Denmark, DK-4000 Roskilde, Denmark; e-mail: [email protected]. Jeppe Seidelin Dam is a postdoctoral student and Peter Tidemand-Lichtenberg is an associate professor at DTU Fotonik.