1. Introduction

Colloidal quantum dots (CQDs) are semiconductor nanocrystals with diameters between 2 to 20 nm [1–4], and they can be synthesized by low-cost chemical methods [1–3,5]. Emission wavelengths of CQDs can be easily tuned from the visible to near infrared band by changing compositions and sizes of the CQDs [5–7]. In addition, CQDs have many excellent properties, such as large absorption, high quantum yields (∼100%), solution processability, stability, and possibility of self-assembly [1,3,5,6]. Therefore, acting as nanoscale building blocks, CQDs have been widely used to construct various nanostructures, which have important applications in the areas of photodetectors [8–11], displayers [12–16], and lasers [6,17–26]. For example, by using the large absorption properties of CQD layers, various CQD-based photodetectors including photodiodes [8], phototransistors [9], infrared photodetectors [10], and dual-band detectors [11] were demonstrated. By employing properties of high quantum yields and tunable emission wavelengths of CQDs, a number of CQD light-emitting diodes (QLED) [12–16] were realized. Besides, CQDs have attracted great attentions in lasing applications due to their large gains in the past two decades. For example, by using CQDs as gain mediums, CQD vertical-cavity surface-emitting lasers (VCSEL) [22], whispering-gallery mode (WGM) lasers [17–20,25], spasers [21,26], and distributed-feedback (DFB) lasers [23,24] were experimentally demonstrated.

Although CQDs have many applications as absorption [8–11] or emission materials [6,12–26], they are seldom used in passive nanophotonic devices, especially in photonic integrated circuits (PICs), where low propagation losses are strongly required. The reason is that CQDs have large absorption losses at wavelengths shorter than their fluorescence wavelengths. In the past decade, enormous efforts have been made to integrate the CQD-based active devices and dielectric-based (or metal-based) passive photonic devices on chips to construct PICs. [17,21,24,27] However, the on-chip integration of the CQD-based active photonic devices and passive photonic devices made of different materials considerably increases the difficulty of fabrication or brings damages to CQDs. [17,21,24,27] For example, the on-chip integrations of the CQD lasers and the SiN waveguides were realized through a complex CMOS-like processes [24,27], including deposition of films, CQD spin-coating, the plasma-enhanced chemical vapor deposition (PECVD) process, and the reactive ion etching (RIE) process. Here, the PECVD process caused a reduction of the CQD photoluminescence (PL) intensity by 80% [27]. It is a possible way to simplify the fabrication and avoid damages to gain materials by using only CQDs to develop full-CQD PICs. However, in PICs with the active devices and passive photonic devices, miscibility of different CQDs can severely destroy the function of isolated CQD devices. As far as we know, combinations of two or more properties of CQDs are not reported.

In this work, passive nanophotonic devices and circuits are experimentally demonstrated by using CQD waveguides at wavelengths longer than their fluorescence wavelengths. At longer wavelengths, CQDs exhibit weak absorptions. By stacking high-index CQDs in a poly(methyl methacrylate) (PMMA) trench [17], CQD waveguides are obtained, and the propagation length of the waveguide mode (λ=800 nm) is measured to be about L_p=1.2 mm, which is more than 40 times that at a short wavelength (λ=630 nm) [28]. Based on the low-loss CQD waveguide, various passive nanophotonic devices are experimentally demonstrated, including Y-splitters, Mach-Zehnder (MZ) interferometers, and directional couplers. We also experimentally demonstrate a low-loss CQD nanophotonic circuit, which integrates waveguides, Y-splitters, MZ interferometers, waveguide-ring resonators (WRRs), and gratings on a single chip. Moreover, a low-loss CQD waveguide and a high-gain CQD laser are experimentally integrated on a single chip by using different CQDs without any mixing. Combining low-loss CQD waveguides and functional devices with CQD lasers and CQD photodetectors [8–11], which are made of different CQDs, it is expected to realize full-CQD PICs.

2. Results and discussion

2.1 Light guiding performance of the CQD waveguides

Waveguides are the fundamental element for the nanophotonic devices and circuits, so light guiding performances of CQD waveguides are firstly experimentally tested at longer wavelengths (>fluorescence wavelength). CQDs are stacked in a straight trench together with two grating trenches on a PMMA film (refractive index n_PMMA≈1.5), which is spin-coated on a glass substrate. The detail fabrication process can be seen in Methods. Herein, the CQDs are quasi-type-II CdSe/ZnSe/ZnS core/shell CQDs [6,17,28] with a diameter of approximately 10 nm. The refractive index of the CQD film is measured to be about n_CQD=1.9 at λ=800 nm by an ellipsometer. Therefore, a CQD strip with a high refractive index could be used as a waveguide. The dark-field optical image of one fabricated CQD straight waveguide (length of L≈30 µm) is shown in Fig. 1(a). At both ends of the waveguide, a grating with a period of 500 nm and a duty cycle of 0.5 is fabricated to couple and decouple free-space light. The sectional view of one CQD waveguide is shown by a scanning electron microscopy (SEM) image in Fig. 1(b). Herein, a gold layer with a thickness of about 130 nm is sputtered above the sample for charge conduction in focused-ion-beam (FIB) etching. It is observed that sidewalls of the PMMA trench are very sharp, and CQDs are compactly stacked in the trench. From Fig. 1(b), the width and height of the CQD waveguide are measured to be approximately w=600 nm and h=370 nm, respectively. Besides, it is observed that there is a thin CQD layer (thickness ∼100 nm) on the PMMA film, as shown in Fig. 1(b).

 

Fig. 1. CQD straight waveguide. (a) Dark-field optical image of a CQD straight waveguide (L=30 µm, scale bar 10 µm). (b) SEM image of the sectional view of a CQD straight waveguide. (c) Cross-section schematic of a CQD straight waveguide (w=600 nm, h=370 nm and t=100 nm). Intensity (|E|²) distributions of the (d) TE₀ waveguide mode and (e) TM₀ waveguide mode. The cyan arrows denote the electric field vectors.

Download Full Size | PDF

Mode properties of waveguides are closely related to structure parameters, so modes supported by CQD waveguides are numerically investigated by the finite element method (FEM) of COMSOL Multiphysics. The cross-section schematic of the CQD waveguide is depicted in Fig. 1(c), which is identical to that of the experimental sample [Fig. 1(b)]. In the simulation, the refractive indices of the glass substrate, PMMA, solid CQD layer, and air are n_sub=1.5, n_PMMA=1.5, n_CQD≈1.9, and n_air=1.0, respectively. The vacuum wavelength is λ=800 nm. According to the simulation results, two transverse electric (TE) waveguide modes and two transverse magnetic (TM) waveguide modes (TE₀, TM₀, TE₁, and TM₁) are supported in the CQD waveguide. The effective refractive indices of these four modes are 1.74, 1.72, 1.56, and 1.54, respectively. The electric intensity distribution (|E|²) and electric field vectors of the TE₀ and TM₀ modes are depicted in Fig. 1(d) and (e), respectively. It is observed that the electric field vectors of the TE mode are mainly parallel to the glass surface, and electric field vectors of the TM mode are mainly vertical to the glass surface. Moreover, the electromagnetic fields are well confined in the CQD waveguide. As a result, removing of the PMMA film and thin CQD layer (thickness ∼100 nm on the PMMA film) is not necessary here, making the fabrication process of CQD waveguides much easier. Unless otherwise specified, CQD nanophotonic devices constructed by the CQD waveguides possess the same lateral dimensions of waveguides (w=600 nm and h=370 nm) in the following experiments.

As we know, propagation properties of waveguides are greatly affected by absorptions of materials in waveguides, so an absorption spectrum of CQDs (dispersed in hexane) is measured, and the results are depicted by the black line in Fig. 2(a). It is observed that CQDs exhibit strong absorptions at wavelengths shorter than ∼520 nm, indicated by the blue shaded area in Fig. 2(a). Besides, CQDs show weak absorptions at near infrared wavelengths (λ=700∼900 nm), as depicted by the red shaded area in Fig. 2(a). Under excitation of a laser beam with a wavelength of λ=430 nm, the measured photoluminescence (PL) spectra of a CQD film (thickness=400 nm) is displayed by the red line in Fig. 2(a). The central peak and full width at half-maximum (FWHM) of the PL spectrum are λ₀≈645 nm and Δλ_FWHM≈50 nm, respectively. At the emission wavelength of λ₀≈645 nm, CQDs still have a small absorption, as denoted by the blue dashed line in Fig. 2(a). However, at λ=800 nm, which is much greater than the emission wavelength (λ₀≈645 nm), the absorption of the CQDs is very small (beyond the detection resolution of our spectrophotometer), and its signal is submerged in noises, as shown by the magenta dashed line in Fig. 2(a). A dark-field optical image of a CQD film is shown in inset of Fig. 2(a), and no defects or pinholes are observed. The SEM and AFM measurements in our previous work [6] showed that CQDs films fabricated by using our drop-casting method were very flat. The surface roughness of the CQD film was about 10 nm [6], which was comparable to the diameters of the CQDs.

 

Fig. 2. Optical properties of the CQDs and propagation lengths of the CQD waveguide. (a) Absorption (black line) and PL (red line) spectra of the CQDs. The blue and magenta dashed lines denote the wavelength of λ=650 nm and λ=800 nm, respectively. Inset is the dark-filed optical image of a CQD film with a thickness of ∼270 nm. The scale bar in the inset is 50 µm. (b) Measured intensities ln(I) of the scattered light from the right gratings of the CQD straight waveguides with different lengths at λ=650 nm (black circle symbols) and λ=800 nm (red square symbols). The black and red solid lines are the linear fitting curves of measured dots of λ=650 nm and λ=800 nm, respectively. The unit a.u. in (a-b) is arbitrary unit. Optical images of the CQD straight waveguides at λ=650 nm with (c) L=20 µm and (d) L=80 µm. The exposure time of the optical image in (d) is twice longer than that in (c). Optical images of the CQD straight waveguides at λ=800 nm with (e) L=20 µm and (f) L=80 µm. The exposure time of the optical images in (e) and (f) is the same.

Download Full Size | PDF

To test light guiding properties of the CQD waveguides at different wavelengths, the propagation length (corresponding to the propagation loss) of the CQD waveguide is experimentally measured. Optical measurements are carried out by a homemade microscope system [29]. When a TE-polarized (electric field vectors perpendicular to waveguides) monochromatic laser beam (λ=800 nm) is focused on the left grating of the waveguide by an objective (100×, NA=0.8), the TE₀ mode of the CQD waveguide is excited due to the polarization and momentum matching. Then, the TE₀ mode propagates along the CQD waveguide until it is scattered by the right grating. The scattered light from the right grating is collected by the same objective and then is imaged on a complementary-metal-oxide-semiconductor (CMOS) camera by a lens with a focal length of f_lens=10 cm.

In order to acquire propagation lengths of CQD waveguides, CQD waveguides with different lengths (L) are fabricated, and the intensities (I) of the scattered light from the right grating are measured. The measurement results are displayed by the black circle symbols in Fig. 2(b). According to the linear fitting curve (black solid line in Fig. 2(b)), the propagation length [30] of the TE₀ waveguide mode (at λ=800 nm) is calculated to be about L_p=1.2 mm. The corresponding optical loss of the straight CQD waveguides is α=10×lg(e)/L_p=10×lg(e)/(0.12)≈36.2 dB/cm. The correlation coefficient R² is 0.30, which is smaller than an ideal value 1. This is attributed to that the propagation length (L_p≈1.2 mm) is much greater than the waveguide lengths (from 10 to 100 µm) used in the experiment. As a result, slight deviations of the focusing position on the grating in each measurement have more influence on the measured intensities than the variations of the waveguide lengths. We also measured the scattered intensities at λ=650 nm, and the results are shown by the red square symbols in Fig. 2(b). By linearly fitting(red solid line in Fig. 2(b)) the red square symbols in Fig. 2(b), the propagation length is calculated to be about L_p=67 µm, which is only about 5.7% of that (L_p≈1.2 mm) at λ=800 nm. This is attributed to that the absorption of the CQDs at λ=650 nm is greater than that at λ=800 nm, as indicated by the blue and magenta dashed lines in Fig. 2(a). Besides, the correlation coefficient R² in the linearly fitting (at λ=650 nm) becomes R²=0.99, which is much closer to 1 than that at λ=800 nm (R²=0.30). This is attributed to that the propagation length (L_p≈67 µm) of the CQD waveguide at λ=650 nm is comparable to waveguide lengths (L=10∼100 µm) in the experiment.

The optical images for waveguides with the length of L=20 µm and L=80 µm at λ=650 nm and λ=800 nm are shown in Fig. 2(c-f), respectively. Here, the exposure time of Fig. 2(d) is twice that of Fig. 2(c). At λ=650 nm, it is observed that the intensity of the scattered light at the right grating for the longer waveguide [L=80 µm in Fig. 2(d)] is much weaker than that for the shorter waveguide [L=20 µm in Fig. 2(c)]. However, at λ=800 nm [see Fig. 2(e) and 2(f)], the scattered intensities from the right grating (red dashed rectangle) for different waveguide lengths (L=20 µm and L=80 µm) are nearly equal. This phenomenon further confirms low losses (corresponding to long propagation lengths) of the CQD waveguides at the long wavelength of λ=800 nm.

At shorter wavelengths (λ<650 nm), the propagation length would become much shorter due to increased absorptions [Fig. 2(a)]. For example, the propagation length of the CQD waveguide at 630 nm becomes about L_p=28 µm [28], which is only about 40% that (L_p≈67 µm) at λ=650 nm. The simulation shows that the CQD waveguide with smaller sectional sizes has longer propagation lengths at the same wavelength. This is attributed to that the proportion of the electric field intensities |E|² confined in the absorptive CQD strip decreases as the sectional sizes decreases. The sizes (w=600 nm and h=370 nm) of the CQD waveguide in this work are greater than that (w=300 nm and h=300 nm) in our previous work [28]. So, at λ=630 nm, the propagation length of the CQD waveguide in this work is smaller than that in our previous work. The ratio of the propagation length of the CQD waveguide in this work (at λ=800 nm) to that in our previous work (at λ=630 nm) is approximately 40. When the geometry structures of the two comparing waveguides are the same, this ratio will become greater.

Bending losses of CQD bending waveguides are also investigated. The schematic of a CQD bending waveguide structure is shown in Fig. 3(a), which is composed of one quarter circle waveguide (radius R) and two straight waveguides (lengths L₁=L₂=20 µm). The dark-field optical image of one fabricated bending waveguide (R=40 µm) on a glass substrate is depicted in Fig. 3(b). When a TE-polarized laser beam is focused on the grating A, a scattered spot is observed at the grating B, as depicted by the red dashed rectangle in Fig. 3(c). At λ=800 nm, the measured intensities of the scattered light from the grating B for different bending radii are shown by the black square symbols in Fig. 3(d). The red solid line in Fig. 3(d) are the smooth curve of the measured data at λ=800 nm. When the radius R is smaller than 15 µm at λ=800 nm, the intensity of the scattered light from the grating B increases rapidly because radiation losses dominate the bending loss [28,29,31]. When the radius is greater than 15 µm, propagation losses starts to dominate the bending loss. The intensity of the scattered light from the grating B remains unchanged when R>15 µm at λ=800 nm owing to the low propagation loss (propagation length L_p≈1.2 mm), as shown by the black square symbols in Fig. 3(d).The experimental result of CQD bending waveguides also confirms the low loss of CQD waveguides at the long wavelength of λ=800 nm.

 

Fig. 3. CQD bending waveguide. (a) Schematic of the CQD bending waveguide. (b) Dark-field optical image of the CQD bending waveguide (R=30 µm). (c) Optical image of the CQD bending waveguide (R=30 µm). The red dashed rectangle denotes the position of the grating B. (d) Measured intensities of the scattered light from the grating B with different bending radius at λ=800 nm (black square symbols). The red solid line is the smoothed curve of the measured dots.

Download Full Size | PDF

2.2 Passive nanophotonic devices with the low-loss CQD waveguide

By using low-loss CQD waveguides, various nanophotonic devices, including Y-splitters, MZ interferometers, and directional couplers, are fabricated and experimentally demonstrated. The schematic of a Y-splitter is shown in Fig. 4(a), which consists of a Y-shaped structure and three straight waveguides. Here, the curved part in the Y-splitter has the cosine–arc shape [32], and it is described as

 

Fig. 4. CQD Y-splitter. (a) Schematic of the CQD Y-splitter. (b) Dark-field optical image of the Y-splitter (W_ys=20 µm and H_ys=6 µm) before drop-casting the CQDs. (c) SEM image of the sectional view of a Y-splitter. A gold film with a thickness of 130 nm is sputtered on the sample for conduction in the etching and SEM processes. Optical images of the CQD Y-splitters with (d) W_ys=20 µm and H_ys=4 µm and (e) W_ys=20 µm and H_ys=6 µm. The red dashed rectangles denote the scattered light from the grating B and grating C. (f) Measured splitting ratios η (I_B: I_C) of the Y-splitters with different W_ys and H_ys.

Download Full Size | PDF

MZ interferometers and directional couplers are also fabricated on a glass substrate and demonstrated by using the low-loss CQD waveguide in the experiment. A MZ interferometer is schematically shown in Fig. 5(a), which is composed of two connected Y-splitters. The lengths of the straight waveguides at the end of the two Y-splitters are L_mz1=L_mz3=10 µm. The lengths of the straight waveguides inserted between the two Y-splitters are L_mz2=7 µm. The distance between centers of the two waveguides is denoted by H_mz. The widths of the curved parts are denoted by W_mz. When a TE-polarized laser beam at λ=800 nm is focused on the left grating, a bright scattered spot from the right grating of the MZ interferometer is observed, as denoted by the red dashed rectangle in Fig. 5(b).The bright scattered light is attributed to the constructive interference of light from the two arms, and it also represents a good stacking uniformity of CQDs in PMMA trenches. The schematic of the directional coupler is shown in Fig. 5(c). The lengths of the waveguides at the left and right part of the structure are both L_dc1=L_dc3=10 µm. The length of the two waveguides in the middle part is L_dc2=23 µm. The width of the gap between the two waveguides in the middle part is G_dc=150 nm. The width and height of each curved part are W_dc1=12 µm and H_dc=2 µm, respectively. When a TE-polarized laser beam at λ=800 nm is focused at the grating B of the directional coupler, a bright scattered spot is observed at the grating C, as shown in Fig. 5(d). Thus, the directional coupler is experimentally demonstrated by using the low-loss and uniform CQD waveguides.

 

Fig. 5. CQD MZ interferometer and directional coupler. Schematic of the (a) CQD MZ interferometer and (c) CQD directional coupler. (b) Optical images of the CQD MZ interferometer with W_ys=20 µm and H_ys=2 µm. The red dashed rectangle indicates the scattered light from the right grating. (d) Optical image of the CQD directional coupler (G_dc=150 nm and L_dc2=23 µm) when the incident laser beam is focused on the grating B. The red dashed rectangles indicate the positions of the four gratings.

Download Full Size | PDF

2.3 Photonic integrated circuit based on the low-loss CQD waveguides

Furthermore, based on the low-loss CQD waveguides and passive CQD nanophotonic devices, a nanophotonic circuit fabricated on a MgF₂ substrate is experimentally demonstrated. Figure 6(a) shows the schematic and structural parameters of the CQD-based nanophotonic circuit, which includes nine straight waveguides, two bending waveguides, one Y-splitter, one MZ interferometer, two WRRs, and five gratings. Here, the lateral width and height of the CQD waveguide are w=300 nm and h=310 nm, respectively. The lateral widths and heights of WRRs are w=800 nm and h=310 nm, respectively. According to Fig. 3(d), when the diameter of the bending waveguide is larger than 26 µm (radius≈13 µm), the bending loss decreases slowly with the increase of the bending diameter. Hence, we fabricate the WRRs with the diameters of 20 µm and 26 µm for comparison here. The size of the nanophotonic circuit is about 65×60 µm², so the long propagation length (∼ mm) of the low-loss CQD waveguide is sufficient for the circuit.

 

Fig. 6. CQD nanophotonic circuit. (a) Schematic and structural parameters of the CQD-based nanophotonic circuit. Optical images of the nanophotonic circuit when the incident light is focused at the (b) grating A and (c) grating C. The red dashed rectangles indicate the scattered light from decoupling gratings. (d) Transmission spectra measured at the grating C, D, and E when the incident light is focused at the grating A. (e) Transmission spectra measured at the grating A and B when the incident light is focused at the grating C. The blue dashed lines in (d) and (e) indicate the center wavelengths of the peaks.

Download Full Size | PDF

When a TE-polarized beam with a wavelength range from 740 nm to 760 nm is focused on the grating A, the TE₀ mode is excited. Then, the TE₀ mode is coupled to WRR₁ at the resonant wavelengths of WRR₁ and propagates in a counter-clockwise direction, as shown by the red arrows in Fig. 6(b). Then, at the resonant wavelength of WRR₂, the TE₀ mode in WRR₁ is coupled to WRR₂ and propagates in a clockwise direction, as shown by the red arrow in Fig. 6(b). At last, the TE₀ mode is coupled to the straight waveguide, and the grating C becomes bright. At the off-resonant wavelengths of WRR₁, the TE₀ mode propagates along the waveguide 1 directly. Then, it transmits to the MZ interferometer and Y-splitter. At last, it is scattered by the gratings D and E, as shown in Fig. 6(b).

In order to further confirm light propagation in the nanophotonic circuit, transmission spectra at the gratings C, D and E are measured, and the results are shown in Fig. 6(d). Due to the small distance (∼ 2 µm) of the grating D and E, fiber end of the spectrometer cannot spatially resolve the scattered spots from the two gratings after imaging. Thus, the spectra of the gratings D and E are not measured separately. It is observed that many narrow resonant peaks appear at the spectrum measured at the grating C. At these resonant wavelengths, transmission dips are observed in the spectrum measured at the grating D and E, as shown by the blue dashed lines in Fig. 6(d). The bandwiths of the peaks and dips are Δλ₁=0.74 nm and Δλ₂=0.44 nm, respectively, corresponding to quality factors of Q₁=λ/Δλ≈1002 and Q₂=λ/Δλ≈1700, respectively. Herein, Q₁<Q₂ is attributed to that the bending loss becomes smaller when the radius increases, as shown in Fig. 3(d). The optical loss of the WRR₁ and WRR₂ are α₁=20πn_efflg(e)/(Qλ) = 20×3.14×1.7×lg(e)/(1002×747×10⁻⁷) ≈619.4 dB/cm and α₂=20πn_efflg(e)/(Qλ) = 20×3.14×1.7×lg(e)/(1700×748×10⁻⁷)≈364.7 dB/cm, respectively. The measured optical loss (α≈36.2 dB/cm) of the CQD waveguide is much smaller than the optical loss of the WRRs because the optical loss of the WRRs includes the propagation loss, radiation loss, and coupling loss. This phenomenon also reveals the low-loss CQD waveguides at long wavelengths. There are two free spectral range (FSR) values of the peaks and dips (magenta arrows in Fig. 6(d), and they are FSR₁=4.46 nm and FSR₂=3.55 nm, corresponding to WRR₁ and WRR₂, respectively. The simulated FSRs for WRR₁ and WRR₂ are FSR₁=4.45 nm and FSR₂=3.45 nm, respectively. The simulated FSRs are very close to the corresponding experimental FSRs. Similarly, when the incident light is focused at the grating C, bright spots can be observed at the gratings A and B, as shown by the red dashed rectangles in Fig. 6(c). Transmission spectra at the grating A and B are also measured, as shown in Fig. 6(e). It is observed that several peaks and dips appear in the spectra, which is similar to that in Fig. 6(d). Therefore, based on the low-loss and uniform CQD waveguides, the passive CQD nanophotonic circuit is experimentally demonstrated.

2.4 Integration of the low-loss CQD waveguide and high-gain CQD laser

It is more promising to integrate the passive and active CQD nanophotonic devices on a single chip. However, combinations of two or more properties of the CQDs are difficult because miscibility of different CQDs can severely destroy functions of isolated CQD devices. Here, by employing both of a pattern-assisted stacking [17,28] and a transfer-printing approach (see Methods), integration of a low-loss CQD waveguide and a high-gain CQD laser on a glass substrate is experimentally realized by using different CQDs without any mixing, as shown by the dark-field optical image in Fig. 7(a). In this fabrication, the high-gain CQD lasers and low-loss waveguides are built on different chips. So, there is not any miscible problem, and functions of the isolated low-loss waveguide and high-gain laser are still maintained. Usually, one or several on-chip coherent sources are needed for a PIC. [33–34] Our transfer-printing approach only takes a few minutes. Therefore, our transfer-printing approach is one of appropriate methods to combine on-chip sources and waveguides.

 

Fig. 7. Integration of the low-loss CQD waveguide and high-gain CQD laser. (a) Dark-field optical image of the integrated structure including a CQD microplate laser and a CQD waveguide. (b) Emission spectra measured at the edge of the CQD laser under different pump densities. (c) Peak intensities and linewidths of the lasing peaks centered at λ=640.7 nm under different pump intensities. The black lines are the linear fitting curves of the peak intensities (at λ=640.7 nm). (d) PL (red solid line) and the absorption spectra (black solid line) of the green CQDs. The blue and magenta dashed lines denote the wavelength of λ_G=530 nm and λ_R=645 nm, respectively. The red shaded spectral range indicates the PL wavelength range of the red CQDs (around λ_R = 645 nm). The unit a.u. in (d) is arbitrary unit. (e) Optical images of the integrated structure under the pump intensity of 1.1P_th. Inset is the zoomed-in optical image of the grating part. The exposure time of the inset is 20 times longer than that of the optical image of the integrated structure. (f) Emission spectra measured at position A (black line) and position B (red line). The position A and position B are indicated by the red dashed circle and yellow dashed circle in (e), respectively.

Download Full Size | PDF

The high-gain CQD laser (red quadrilateral microplate in Fig. 7(a)) is made of red CdSe/ ZnSe/ZnS core/shell CQDs [6] with a PL peak of λ_R=645 nm (Fig. 2(a)), and its lasing wavelengths are around λ=640 nm under the pump of a picosecond laser of λ=532 nm, as shown in Fig. 7(b). Here, the thickness of the CQD microplate laser is about 610 nm. More details about the CQD microplate laser can be found in our previous work [6]. Figure 7(c) depicts a pump threshold and linewidth narrowing, confirming the lasing action [6,17,28]. The pump threshold and linewidth are P_th=393 µJ cm⁻² and Δλ=0.19 nm, respectively. The low-loss CQD waveguide (green line in Fig. 7(a)) is made of green CdSe/ZnS core/shell CQDs (diameters ∼6 nm) with a PL peak of λ_G=530 nm [blue line in Fig. 7(d)]. The green CQDs exhibit low absorptions at the wavelength of λ≈640 nm, which is close to lasing wavelengths of the red CQD microplate laser, as shown by the black line in Fig. 7(d). The lasing signal from the red CQD microplate can be evanescently coupled to the green CQD waveguide, and the right grating becomes bright, as shown by the inset in Fig. 7(e). Complete overlapping of the measurement spectra from the red CQD microplate laser and the grating at the end of the green CQD waveguide also verifies this coupling, as displayed in Fig. 7(f). There are no technical difficulties in manipulating a CQD microplate lasing and low-loss waveguides at other wavelengths. Therefore, integration of the high-gain CQD laser and low-loss CQD waveguide on a single chip is successfully realized by using different CQDs. It is hopeful to realize full-CQD PICs including the passive and active CQD nanophotonic devices.

3. Conclusions

In summary, nanophotonic devices and circuits were experimentally demonstrated by using low-loss and uniform CQD waveguides at wavelengths longer than fluorescence wavelengths of CQDs. Due to weak absorptions of CQDs at long wavelengths, propagation length of CQD waveguides were measured to be approximately L_p=1.2 mm at λ=800 nm, which was more than 40 times that (L_p∼28 µm) at the short wavelength (λ=630 nm) [28]. By using the low-loss CQD waveguides, various passive nanophotonic devices, including bending waveguides, Y-splitters, MZ interferometers, and directional couplers, were experimentally demonstrated. Especially, in CQD Y-splitters, the average of the splitting ratios (η=0.94) was close to an ideal value (η=1), indicating a good uniformity of the CQD waveguides. Based on these functional CQD devices, a nanophotonic circuit was successfully demonstrated. Furthermore, by employing both of a pattern-assisted stacking and a transfer-printing approach, the miscible problem of different CQDs was addressed for the first time, and a low-loss CQD waveguide and a high-gain CQD laser were experimentally integrated on a single chip. This method not only simplified PICs’ fabrications but also avoided CQD damages. Therefore, CQDs could be used as not only absorption and emission materials but also low-loss materials. By using different optical properties of CQDs at different wavelengths, it is hopeful to realize full-CQD PICs including on-chip lasers (around fluorescence wavelengths), low-loss waveguides (greater than fluorescence wavelengths), and detectors (shorten than fluorescence wavelengths).

4. Methods

4.1 CQD waveguides, passive nanophotonic devices, and nanophotonic circuits

The pattern-assisted stacking approach [17,28] is used to fabricate CQD waveguides, passive nanophotonic devices, and CQD PICs. Firstly, trench patterns are fabricated by a high-resolution electron beam lithography (EBL, Raith e-LINE plus) on a 370-nm-thick PMMA film (AR-P 672.045, Allresist). Secondly, quasi-type-II CdSe/ZnSe/ZnS core/shell quantum dots (Wuhan Jiayuan Quantum Dots Co., Ltd., China) dispersed in a mixed solvent (octane: hexane ∼ 1:5) with a concentration of 30 mg/mL are drop-casted on the PMMA film. After evaporation of the mixed solvent, CQDs are stacked into the designed trenches.

4.2 CQD microplate laser and transfer printing

CQD microplate lasers are built by a mixed solvent of CQDs on another chip. The detailed fabrication method can be found in our previous work [6]. The combination of a CQD laser and a CQD waveguide is as follows, and no mixing of different CQDs occurs. Firstly, a target sample with low-loss CQD waveguides is equipped on a rotatable translation stage under a dark-field microscope. Secondly, CQD microplate lasers on a PDMS stamp (∼4 mm×4 mm) are equipped on a rotatable three-axis translation stage, and the PDMS stamp are located directly above CQD waveguides. Thirdly, a CQD microplate laser and a CQD waveguide are aligned under the dark-field microscope by the translation stages. Fourthly, the PDMS stamp is landed on the surface of the target sample. At last, the CQD microplate laser is transferred on the CQD waveguide by lifting the PDMS stamp. The reason is that the adhesion between the CQD microplate laser and the surface of the target sample is larger than that between the CQD microplate laser and the PDMS. After the stamp transfer-printing process, the wavelength of the dominant lasing peak shows a slight red shift (from 639.7 nm to 640.7 nm), and the threshold of the microplate laser becomes slightly higher (from P_th=357 µJ cm⁻² to P_th=393 µJ cm⁻²). This is attributed to that the substrate is changed after the stamp transfer-printing process.

4.3 Measurement of the CQD microplate laser

Lasing characteristics of CQD microplate lasers are measured by a homemade microscope system [17]. A CQD microplate laser is irradiated by a picosecond pump laser, and the wavelength, pulse duration, and repetition rate of the pump laser are 532 nm, 200 ps, and 1 kHz, respectively. The pump beam is focused onto a sample from the substrate side by a quartz lens with a focal length of f=35 mm. The pump spot size is approximately 200 ×200 µm². Then, the emission light from CQD microplate lasers is collected by an objective (Olympus 20×, NA 0.4). Next, lasing emission is divided into two paths and collected by a CMOS camera and a spectrograph, respectively. Here, a long pass filter is used to filter the pump light before a beam splitter.