We report a comparison of different capping materials on the intermixing of modulation p-doped InAs/In(Ga)As quantum dots (QD). QD materials with different caps are shown to exhibit significant difference in their optical properties during the annealing process. The selective area intermixing technique is demonstrated to laterally integrate two and three different QD light emitting devices with a single electrical contact. A spectral bandwidth of 240nm centered at 1188nm is achieved in a device with two sections. By calculating the point spread function for the obtained emission spectra, and applying the Rayleigh criteria for resolution, an axial resolution of 3.5μm is deduced. A three section device realizes a spectral bandwidth of 310nm centered at 1145nm. This corresponds to an axial resolution of 2.4μm. Such a small predicted axial resolution is highly desirable in optical coherence tomography system and other coherence-based systems applications.
© 2012 OSA
Broadband light sources are vital components for wavelength division multiplexing (WDM) , fibre optic gyroscopes (FOG) , and optical coherence tomography (OCT) systems . The shape of the spectrum is of great importance in determining the axial resolution in the OCT system. The wide spectral bandwidth of the light source determines the coherence length which in turn determines the resolution of imaging . In the past ~10 years, self-assembled quantum dot (QD) structures have attracted considerable attention for the realization of broadband optical sources due to their inhomogeneously broadened emission spectra [5, 6]. Various methods have been proposed and utilized to achieve broad spectral bandwidth light emission from QD devices, such as using multilayer stacks of QDs with different emission wavelength for each layer , hybrid quantum well (QW)/quantum dot structures , optimizing the growth conditions to increase the inhomogeneous dot size distribution , or using multi-contact device structures . In addition to QD epitaxial growth and device fabrication techniques, most recently, post-growth intermixing processes are emerging as a very effective method to broaden the emission spectrum , by increasing the effect of interface fluctuations between the QDs and their surrounding barrier layer materials . Subsequently, a selective area post-growth intermixing technique has been successfully demonstrated to laterally integrate two different optical elements (quantum dot superluminescent diodes) by realizing a spatial variation of the bandgap energy of quantum dot materials across a single wafer. However, due to a strong spectral overlap of the two different regions, the selective area intermixed device did not exhibit enhanced emission bandwidth .
Post-growth annealing processes have been investigated for ~30 years, and have been widely used to modify the optical properties of semiconductor materials and devices. Selective area QW intermixing (QWI) techniques have been used for the fabrication of photonic integrated circuits (PICs) , and have also been used to realize high power single mode lasers by introducing non-absorbing mirror facets . More recently, quantum dot intermixing (QDI) has been used to fabricate wavelength tunable QD lasers  and passive QD devices [17, 18]. QDI is more complicated than QWI because the intermixing process is not only influenced by the difference in thermal expansion coefficients between the QDs and the surrounding materials, but is also strongly affected by the shape, size and strain distribution in/around the QDs . As a result there are fewer reports in the literature on QDI compared to QWI. Furthermore, the high thermal sensitivity of QD structures makes the optimized annealing parameters (e.g. annealing temperature and time duration) very difficult to trace.
Recently, introducing p-type doped QD active materials has been demonstrated to improve the thermal stability of QDs during the annealing process, leading to the realization of high performance active intermixed QD devices [13, 20]. Several different capping layers have been used during the annealing process, such as TiO2 , aluminium (Al)  (which have been used to reduce the interdiffusion rate) and SiO2 (which can enhance the interdiffusion rate) . Thin TiO2 films have been demonstrated  to significantly reduce the intermixing degree due to its large thermal expansion coefficient. When a thin dielectric film is deposited onto a thick GaAs substrate , the thermal expansion coefficient mismatch between the film and the substrate at the annealing temperature will cause a thermal stress near the interface region during the annealing process. The thermal expansion coefficient of TiO2 is ~8.2 × 10-6 oC−1 which is larger than that of the GaAs (~6.8 × 10-6 oC−1). During the annealing process TiO2 is under compressive stress near the interface region when GaAs is under tensile stress, the Ga vacancies will be trapped which inhibits interdiffusion process. Moreover, Al has a much larger thermal expansion coefficient (~23.5 × 10-6 oC−1) , so it is also expected to inhibit intermixing by the same argument. And also Al capping has been used to reduce the intermixing rate, with this being discussed in terms of an increased density of Ga interstitials on the sample surface due to atomic exchange between the Al film and Ga2O3 layer during the annealing process . Such interstitials are expected to inhibit the formation of Ga vacancies which drive the intermixing process. PECVD deposited SiO2 layer (thermal expansion coefficient lower than GaAs) behaves like a porous structure during the annealing process. Ga atoms readily diffuse into the SiO2 matrix, leaving a number of Ga vacancies in the semiconductor structure which will enhance the degree of interdiffusion. Furthermore, the intermixing rate is found to be increased with increasing the thickness of the SiO2 films. Since there is always a limit of Ga solubility in the SiO2 film, once it reaches this value, no more Ga vacancies can be generated. The solubility is higher in thicker SiO2 film (e.g. 500nm film). Moreover, related to the mismatch of thermal expansion, the stress caused at the SiO2/GaAs interface is larger with a thicker SiO2 film, which also enhances the Ga atoms out-diffusion into the SiO2 matrix .
In this paper, a comparative study of the intermixing of modulation p-doped QD structures using various caps to promote and inhibit intermixing is made using the same annealing process. Samples with TiO2 and SiO2 caps show significant difference in emission wavelength, which we utilizing in a selective area intermixing (SAI) process to realize an ultra broadband light source. Two and three different QD light emitting devices are successfully integrated laterally by the selective area intermixing process, with a 240nm broadband emission achieved from an intermixed QD device with TiO2 and SiO2 caps, and an emission of 310nm achieved from an intermixed device with a TiO2 cap and two different SiO2 caps. We go on to assess the devices for interferometric applications such as OCT. By calculating the point spread function (PSF) from the emission spectrum and applying a Rayleigh criterion correction, a maximum axial resolution of ~3.5μm is deduced for the 240nm wide spectrum and an axial resolution of ~2.4μm is similarly deduced for the 310nm wide spectrum.
A 5 layer InAs dot-in-well (DWELL) structure was grown in a molecular beam epitaxy (MBE) Veeco Gen 2 system on a Si-doped GaAs (100) substrate. In each QD layer, 3 monolayer (ML) of InAs is grown on a 2nm In0.15Ga0.85As layer, covered by a 5nm In0.15Ga0.85As layer. The 5 DWELL structures were separated by 44nm GaAs spacers. Modulation p-doping with beryllium (Be) to a concentration of 20 acceptors per dot was located in the 9nm wide GaAs spacer layer, 6nm beneath each DWELL. The whole QD active region was sandwiched by lower n-Al0.4Ga0.6As and upper p-Al0.4Ga0.6As cladding layers, which is shown schematically in the inset of Fig. 1 .
The annealing process was performed in an N2 ambient at a temperature of 700°C for 5mins. These conditions are based on various experiments to achieve large differential shifts in emission wavelength, and limited change in emission intensity. Optimized windows for annealing temperature are 700°C ± 20°C and for duration are 5 ± 1min. Four different kinds of metal and dielectric caps were initially trialed including a 200nm thick e-beam evaporated TiO2 film, a 200nm thick e-beam evaporated aluminum (Al) film, a 200nm thick plasma-enhance chemical-vapor deposition (PECVD) deposited SiO2 film, and a 500nm thick PECVD deposited SiO2 film. Samples for photoluminescence (PL) measurement were prepared by etching off ~1μm of the p-AlGaAs cladding on top of the QD samples. The room temperature PL (RT-PL) result was obtained via excitation using a diode-pumped solid-state laser emitting at 532nm and detected with a Ge Detector. The selective area intermixed QD samples were fabricated into 5μm wide ridge waveguide devices by a dry etch through the QD active region. A thin layer of Au-Zn-Au and In-Ge-Au was thermally evaporated on the top and the bottom of the device to provide p and n-side ohmic contacts, respectively. The waveguide structure is 7 degree off from normal to the facet. 6mm long as cleaved devices were mounted on ceramic tiles and tested at room temperature under pulsed operation with 5μs pulse width and 5% duty cycle to reduce the effect of self-heating.
Figure 1 shows the RT-PL spectra of samples annealed with different caps under an excitation laser power density of 50W/cm2. It can be seen that for the as-grown sample the ground state (GS) emission is located at ~1285nm, the TiO2 capped sample and the Al capped sample have similar GS emission wavelength (at ~1275nm and ~1285nm respectively). Compared to the as-grown sample, there is a very small blue-shift of the GS emission peak wavelength for the TiO2 capped sample and the Al capped sample. This confirms that the TiO2 and Al cap provide an effective way to suppress the inter-diffusion between the QDs and the surrounding matrix during this annealing process. Moreover, for the 200nm SiO2 capped sample, the GS peak blue shifted to 1210nm after annealing and has lower integrated PL intensity. For the 500nm SiO2 capped sample, the GS peak is further blue shifted to ~1089nm, also its PL integrated intensity drops by a factor of ~2.
The room temperature power-dependent PL of the as-grown QD sample, TiO2 capped intermixed QD sample, Al capped intermixed sample, 200nm and 500nm SiO2 capped intermixed QD samples are shown in Fig. 2 . For the as-grown sample, at low pump power (~10W/cm2) the GS is located at ~1285nm, the full width at half maximum (FWHM) of the GS is ~41nm. At higher power (~20W/cm2), the GS saturates due to state filling  and we begin to observe the excited state (ES) at ~1202nm with FWHM of ~58nm. At the highest power (~650W/cm2) the second excited state (ES2) starts to appear at ~1146nm with FWHM of ~68nm. The PL data of the TiO2 capped sample annealed at 700°C is plotted in Fig. 2(b). At ~10W/cm2 the GS is slightly blue-shifted (10nm) to ~1275nm compared to the as-grown one, the FWHM of the GS is ~42nm. At ~20W/cm2, the GS saturates and we begin to observe the ES at ~1191nm with FWHM of ~63nm. At ~650W/cm2 ES2 starts to appear at ~1134nm with FWHM of ~68nm. It has been predicted  that by using a TiO2 cap the long wavelength emission of QDs can be preserved during the SAI process, which is very important for broadband emission light source fabrication. As expected, the spectral shape is very similar to that of the as-grown sample and the PL integrated intensities are essentially the same at the maximum pump power density. The result for the Al capped sample is shown in Fig. 2(c), at ~10W/cm2 the GS is located at ~1285nm and the FWHM of the GS is ~43nm. At ~20W/cm2, the GS saturates and we begin to observe ES at ~1211nm with FWHM of ~74nm. At ~650W/cm2 the ES2 starts to appear at ~1140nm with FWHM of ~80nm. A reduction in ES-GS splitting has been observed in the early stage of QD intermixing . In Fig. 2(d), the spectra of the sample annealed with 200nm SiO2 is plotted, at ~10W/cm2 the GS exhibits a large blue-shift of 75nm (from 1285nm to 1210nm), the FWHM of the GS is ~124nm. There is no second peak appearing with increasing excitation power. The apparent blue-shift of the peak can be attributed to state filling effects when the split in energy is smaller than the inhomogeneous broadening.also, the enhanced interdiffusion observed for SiO2 capped QD samples is in a good agreement with previous reports . At ~650W/cm2 the GS and ES combined peak is located at 1182nm with FWHM of ~132nm. Fig. 2(e) shows the annealed sample capped with a 500nm SiO2 film, at ~10W/cm2 its GS emission peak blue shifted from ~1285nm to ~1087nm (198nm) with FWHM of ~42nm, and the PL integrated intensity drops to ~50% of the as grown sample. There is no significant blue-shift with increasing pump power, this may due to a transition from QD-like to more QW-like behavior with increasing degree of intermixing. Furthermore, the reduced PL intensity is suggestive of a reduced carrier lifetime which may also act to inhibit state-filling effects. This reduced carrier lifetime may be due to reduced confinement and enhanced thermal escape and non-radiative recombination in the GaAs matrix, and/or increased defect related non-radiative recombination in the QDs. At ~650W/cm2 the GS is located at ~1090nm with the FWHM of ~58nm. To a greater or lesser degree, a feature is observed in all spectra at ~990nm. This is attributed to recombination in the QW in which the QDs are encapsulated. It is observed that this emission becomes more pronounced as the confinement energy of the QDs is reduced.
Although an Al mask is a good candidate to preserve the emission wavelength and intensity of the QDs during the annealing treatment, a very rough surface was observed after the Al was removed. Further processing of the sample was not pursued. A solution to this has been proposed however, where a thin dielectric layer between the GaAs and Al is used . The other three capping techniques were investigated further as a smooth sample surface was maintained, ideal for device fabrication.
Firstly, a device was fabricated containing two light emitting elements realized using both a TiO2 cap and a 200nm SiO2 cap. A 5 minute, 700°C anneal was then applied to this sample to give two regions with different emission wavelengths. A 5μm wide ridge waveguide device was fabricated as described previously. The schematic of the device is plotted as an inset in Fig. 3(a) , the TiO2 capped section is chosen to be 2mm and the SiO2 capped section is chosen to be 4mm. The EL spectrum of the device is shown in Fig. 3(a) with a maximum ex-facet power of 2.25mW at a drive current density of 3.67kA/cm2. An emission spectrum of 240nm 3dB band-width centered at ~1188nm is obtained at this maximum power. As there is a large differential shift of the emission wavelengths of the two intermixed QD regions, a broader spectrum is obtained compared to our previous work in , where the spectra from the two intermixed regions were strongly overlapped.
With a light source with Gaussian spectral shape, the coherence length lcFWHM of an OCT system can be theoretically represented as :28]. Based on Eq. (1), lcFWHM /2 is ~2.6μm if the experimentally determined value of 240nm is used. Such resolution may be used to visualize the details of e.g. the human skin tissues, retina, and choroid where features have typical scales between 10 and 30μm [29, 30]. However, Eq. (1) is only satisfied when the spectrum is a pure Gaussian shape. For more complex emission spectra the resolution of an OCT system may be estimated through the complex temporal coherence function (CTCF) which can be regarded as the PSF of the imaging system .
Figure 3(b) plots the modulus of the CTCF for the spectrum in Fig. 3(a) and a Gaussian spectrum with ∆λ = 240nm, obtained using an inverse fast Fourier transform. For the Gaussian spectrum, as expected we obtain a system resolution similar to the calculated one in Eq. (1). For the experimental data, side-lobes are observed in the complex temporal coherence function and half of the FWHM of the peak is 0.010492ps corresponding to ~3.1μm. However, side-lobes in the CTCF close to the main lobe will act to decrease image resolution. The impact of the side-lobes in resolution is explored in Fig. 3(c), where our experimentally obtained CTCF is applied to two layers separated by ∆z (the optical path-length difference). Here, in order to achieve the Rayleigh criterion for resolution ∆z is ~3.5μm. The non-Gaussian emission spectrum of the device may therefore be considered to introduce a 0.4μm penalty to system axial resolution. In addition, ~35% of the energy is lost in the side-lobe.
In order to further increase the −3dB bandwidth of the light emission spectrum, a three region intermixed device was fabricated. This consisted of a 200nm TiO2 film covered region, a 200nm SiO2 film covered region and a 500nm SiO2 film covered region incorporated in one device during the annealing process at 700°C. The EL spectrum of the device is shown in Fig. 3(d) with a maximum ex-facet power of 1.81mW at drive current density of 3.5kA/cm2. A spectrum with 310nm −3dB bandwidth centered at ~1145nm is obtained at this maximum power. As expected, from the PL data presented previously, the spectrum from the TiO2 capped section and the spectra from the other two SiO2 capped sections overlap to form an emission spectrum with a flat top. Figure 3(e) plots the CTCF for experimental data in 3(d) and a Gaussian spectrum with ∆λ = 240nm. Half of the FWHM of experimental system equals to 0.007187ps correspond to ~2.2μm. Again the effect of the side-lobes on axial resolution is explored in 3(f), where our experimentally obtained CTCF is applied to two layers separated by ∆z. Here, in order to achieve the Rayleigh criterion the axial resolution is ~2.4μm. This gives us a corrected axial resolution of 2.4μm for the three sections intermixed device. These theoretical values for OCT system resolution are promising for sub-cellular resolution imaging.
In summary, we directly compare the effect of the different caps on the optical properties of QDs with a view to the suitability of these caps to SAI of active QD device. By fabricating devices with suitable caps and choice of lengths, devices with two and three differentially intermixed regions are realized. FWHMs of 240nm and 310nm with ex-facet mW power levels are obtained for the two and three intermixed region devices, respectively. The practical system axial resolution possible, performing OCT with these devices is discussed with resolutions of 3.5μm and 2.4μm being deduced using the Rayleigh criterion, for the two and three intermixed region devices, respectively.
This work was supported by EPSRC grant EP/I018328/1
References and links
1. R. D. Feldman, E. E. Harstead, S. Jiang, T. H. Wood, and M. Zirngibl, “An evaluation of architectures incorporating wavelength division multiplexing,” J. Lightwave Technol. 16(9), 1546–1559 (1998). [CrossRef]
2. W. Burns, C. Lin, and R. Moeller, “Fiber-optic gyroscopes with broad-band sources,” J. Lightwave Technol. 1(1), 98–105 (1983). [CrossRef]
3. W. Drexler, U. Morgner, F. X. Kärtner, C. Pitris, S. A. Boppart, X. D. Li, E. P. Ippen, and J. G. Fujimoto, “In vivo ultrahigh-resolution optical coherence tomography,” Opt. Lett. 24(17), 1221–1223 (1999). [CrossRef] [PubMed]
5. Z. Y. Zhang, Z. G. Wang, B. Xu, P. Jin, Z. Z. Sun, and F. Q. Liu, “High-performance quantum-dot superluminescent diodes,” IEEE Photon. Technol. Lett. 16(1), 27–29 (2004). [CrossRef]
6. Z. Y. Zhang, R. A. Hogg, X. Q. Lv, and Z. G. Wang, “Self-assembled quantum-dot superluminescent light-emitting diodes,” Adv. Opt. Photon. 2(2), 201–228 (2010). [CrossRef]
7. L. H. Li, M. Rossetti, A. Fiore, L. Occhi, and C. Velez, “Wide emission spectrum from superluminescent diodes with chirped quantum dot multilayers,” Electron. Lett. 41(1), 41–43 (2005). [CrossRef]
8. S. M. Chen, K. J. Zhou, Z. Y. Zhang, D. T. D. Childs, M. Hugues, A. J. Ramsay, and R. A. Hogg, “Ultra-broad spontaneous emission and modal gain spectrum from a hybrid quantum well/quantum dot laser structure,” Appl. Phys. Lett. 100(4), 041118 (2012). [CrossRef]
9. X. Q. Lv, N. Liu, P. Jin, and Z. G. Wang, “Broadband emitting superluminescent diodes with InAs quantum dots in AlGaAs matrix,” IEEE Photon. Technol. Lett. 20(20), 1742–1744 (2008). [CrossRef]
10. Y. C. Xin, A. Martinez, T. Saiz, A. J. Moscho, Y. Li, T. A. Nilsen, A. L. Gray, and L. F. Lester, “1.3μm quantum-dot multisection superluminescent diodes with extremely broad bandwidth,” IEEE Photon. Technol. Lett. 19(7), 501–503 (2007). [CrossRef]
11. Q. Jiang, Z. Y. Zhang, M. Hopkinson, and R. A. Hogg, “High performance intermixed p-doped quantum dot superluminescent diodes at 1.2μm,” Electron. Lett. 46(4), 295–296 (2010). [CrossRef]
12. Z. Y. Zhang, R. A. Hogg, B. Xu, P. Jin, and Z. G. Wang, “Realization of extremely broadband quantum-dot superluminescent light-emitting diodes by rapid thermal-annealing process,” Opt. Lett. 33(11), 1210–1212 (2008). [CrossRef] [PubMed]
13. Z. Y. Zhang, Q. Jiang, M. Hopkinson, and R. A. Hogg, “Effects of intermixing on modulation p-doped quantum dot superluminescent light emitting diodes,” Opt. Express 18(7), 7055–7063 (2010). [CrossRef] [PubMed]
14. J. H. Marsh, “Quantum well intermixing,” Semicond. Sci. Technol. 8(6), 1136–1155 (1993). [CrossRef]
15. C. L. Walker, A. C. Bryce, and J. H. Marsh, “Improved catastrophic optical damage level from laser with nonabsorbing mirrors,” IEEE Photon. Technol. Lett. 14(10), 1394–1396 (2002). [CrossRef]
16. H. S. Djie, Y. Wang, D. Negro, and B. S. Ooi, “Postgrowth band gap trimming of InAs/InAlGaAs quantum-dash laser,” Appl. Phys. Lett. 90(3), 031101 (2007). [CrossRef]
17. A.-R. Bellancourt, Y. Barbarin, D. J. H. C. Maas, M. Shafiei, M. Hoffmann, M. Golling, T. Südmeyer, and U. Keller, “Low saturation fluence antiresonant quantum dot SESAMs for MIXSEL integration,” Opt. Express 17(12), 9704–9711 (2009). [CrossRef] [PubMed]
18. Z. Y. Zhang, A. E. H. Oehler, B. Resan, S. Kurmulis, K. J. Zhou, Q. Wang, M. Mangold, T. Suedmeyer, U. Keller, K. J. Weingarten, and R. A. Hogg, “1.55μm InAs/GaAs quantum dots and high repetition rate quantum dot SESAM mode-locked laser, ” Sci. Rep. 2, Article Nr. 477 (2012).
19. X. C. Wang, S. J. Xu, S. J. Chua, Z. H. Zhang, W. J. Fan, C. H. Wang, J. Jiang, and X. G. Xie, “Widely tunable intersubband energy spacing of self-assembled InAs/GaAs quantum dots due to interface intermixing,” J. Appl. Phys. 86(5), 2687–2690 (1999). [CrossRef]
20. Z. Y. Zhang, Q. Jiang, and R. A. Hogg, “Tunable interband and intersubband transitions in modulation C-doped InGaAs/GaAs quantum dot lasers by postgrowth annealing process,” Appl. Phys. Lett. 93(7), 071111 (2008). [CrossRef]
21. L. Fu, P. Lever, H. H. Tan, C. Jagadish, P. Reece, and M. Gal, “Suppression of interdiffusion in InGaAs/GaAs quantum dots using dielectric layer of titanium dioxide,” Appl. Phys. Lett. 82(16), 2613–2615 (2003). [CrossRef]
22. R. M. Cohen, G. Li, C. Jagadish, P. T. Burke, and M. Gal, “Native defect engineering of interdiffusion using thermally grown oxides of GaAs,” Appl. Phys. Lett. 73(6), 803–805 (1998). [CrossRef]
23. A. Pepin, C. Vieu, M. Schneider, H. Launois, and Y. Nissim, “Evidence of stress dependence in SiO2/Si3N4 encapsulation-based layer disordering of GaAs/AlGaAs quantum well heterostructures,” J. Vac. Sci. Technol. B 15(1), 142–153 (1997). [CrossRef]
24. S. Alexey, “Properties of pure aluminum, ” in Handbook of Aluminum (CRC Press, 2003), Chap. 2.
25. B. S. Ooi, K. McIlvaney, M. W. Street, A. S. Helmy, S. G. Ayling, A. C. Bryce, J. H. Marsh, and J. S. Roberts, “Selective quantum-well intermixing in GaAs-AlGaAs structures using impurity-free vacancy diffusion,” IEEE J. Quantum Electron. 33(10), 1784–1793 (1997). [CrossRef]
26. S. Grosse, J. H. H. Sandmann, G. von Plessen, J. Feldmann, H. Lipsanen, M. Sopanen, J. Tulkki, and J. Ahopelto, “Carrier relaxation dynamics in quantum dots: scattering mechanisms and state-filling effects,” Phys. Rev. B 55(7), 4473–4476 (1997). [CrossRef]
27. Z. Y. Zhang, Q. Jiang, I. J. Luxmoore, and R. A. Hogg, “A p-type-doped quantum dot superluminescent LED with broadband and flat-topped emission spectra obtained by post-growth intermixing under a GaAs proximity cap,” Nanotechnology 20(5), 055204 (2009). [CrossRef] [PubMed]
28. Y. Zhang, M. Sato, and N. Tanno, “Numerical investigations of optimal synthesis of several low coherence sources for resolution improvement,” Opt. Commun. 192(3-6), 183–192 (2001). [CrossRef]
29. Y. H. Zhao, Z. P. Chen, C. Saxer, S. H. Xiang, J. F. de Boer, and J. S. Nelson, “Phase-resolved optical coherence tomography and optical Doppler tomography for imaging blood flow in human skin with fast scanning speed and high velocity sensitivity,” Opt. Lett. 25(2), 114–116 (2000). [CrossRef] [PubMed]
30. W. Drexler, U. Morgner, R. K. Ghanta, F. X. Kärtner, J. S. Schuman, and J. G. Fujimoto, “Ultrahigh-resolution ophthalmic optical coherence tomography,” Nat. Med. 7(4), 502–507 (2001). [CrossRef] [PubMed]