Graphene-Quantum-Dot-Mediated Semiconductor Bonding: A Route to Optoelectronic Double Heterostructures and Wavelength-Converting Interfaces

A semiconductor bonding technique that is mediated by graphene quantum dots is proposed and demonstrated. The mechanical stability, electrical conductivity, and optical activity in the bonded interfaces are experimentally verified. First, the bonding scheme can be used for the formation of double heterostructures with a core material of graphene quantum dots. The Si/graphene quantum dots/Si double heterostructures fabricated in this study can constitute a new basis for next-generation nanophotonic devices with high photon and carrier confinements, earth abundance, environmental friendliness, and excellent optical and electrical controllability via silicon clads. Second, the bonding mediated by the graphene quantum dots can be used as an optical-wavelength-converting semiconductor interface, as experimentally demonstrated in this study. The proposed fabrication method simultaneously realizes bond formation and interfacial function generation and, thereby, can lead to efficient device production. Our bonding scheme might improve the performance of optoelectronic devices, for example, by allowing spectral light incidence suitable for each photovoltaic material in multijunction solar cells and by delivering preferred frequencies to the optical transceiver components in photonic integrated circuits.


Introduction
Quantum dots, often referred to as artificial atoms, exhibit extraordinary electronic and optical properties, owing to the three-dimensional confinement of electrical carriers and the discretized density of states [1]. Hence, they enable various applications, including high-performance optoelectronic devices [2][3][4][5][6][7][8] and single-electron manipulation [9][10][11][12]. The graphene quantum dot (GQD) [13][14][15] is a quantum dot that originates from carbon, which is a low-cost, earth-abundant, and environmentally friendly material. In a manner that is similar to other quantum dots, GQD is known to exhibit various electrical and optical characteristics that are based on its size and utilized for various applications [13][14][15][16][17]. However, GQD materials are used independently in the devices as reported to date, and such devices suffer from considerable electrical and optical losses, due to the carrier recombination and photon scattering at the surfaces and interfaces. Double heterostructures [18,19] that permit carrier and photon confinements are employed for practical optoelectronic devices to address this issue [5,8,20].
Semiconductor wafer bonding is a useful fabrication method that is employed in various electronic and photonic applications [21][22][23]. The bonding technique is used to form heterostructures of dissimilar semiconductor materials with high crystalline qualities, while the conventional epitaxial growth method inevitably generates substantial levels of defect densities due to crystalline lattice mismatches.

Materials and Methods
We used a commercially available aqueous solution of colloidal GQDs with a concentration of 0.1 w/v % (GS Alliance Co., Ltd., Kawanishi, Japan) in the study. Silicon, which is a versatile semiconductor, was adopted as the cladding material. We investigate two approaches as the methods to introduce interfacial GQDs: (1) bond formation via directly coating the colloidal GQD solution on the semiconductor wafers and (2) bonding via GQDs that are embedded in a hydrogel material. We performed all of the experimental processes in the study in a non-cleanroom, regular experimental room with a particle density of approximately five-million m −3 , which we measured with a regular particle counter. We used epi-ready-grade, single-side-polished p-type Si wafers (thickness: 280 µm, crystalline plane orientation: <100>, dopant: boron, doping concentration:~1 × 10 19 cm −3 ) for most of the experiments in this study, unless noted otherwise. The surfaces of the polished sides of the Si wafers were first coated with a photoresist film to protect the Si surfaces from scratches and particulates during the process of being diced into approximately 1-cm 2 pieces. The diced wafers were submerged in acetone for five min. to remove the photoresist film and degrease the Si surfaces. Some of the Si pieces were then subjected to a wet hydrofluoric (HF) treatment (10% aq, 1 min.) to remove the SiO 2 native oxide layer formed on the Si surfaces. In some experiments, we used a hydrogel material as an adhesive and viscous organic matrix to embed the GQDs. In this case, a 2, 4, or 6 w/v % polyacrylamide (PAM) aqueous solution was prepared via mixing PAM powder with deionized water and adequately stirring in order to prevent the aggregation of the adhesive PAM particles. Equal volumes of GQD aq. and PAM aq. were mixed, which resulted in a 0.05 w/v % GQD hydrogel (PAM: 1, 2, or 3 w/v %). The bare GQD aq. or prepared hydrogel containing GQDs was then coated onto the polished-side surface of a Si piece. The GQD-coated Si piece was then bonded to the surface of the polished side of a bare Si piece under a uniaxial pressure of 0.1 MPaG in ambient air for 3 h. The bonding experiments were performed at various temperatures in the range of 20-500 • C, and only at room temperature for the GQD-only and hydrogel-mediated cases, respectively. The heating and cooling rates were approximately 10 • C/min.
After the bonding, detachment normal stresses were measured for the bonded samples in order to represent bonded interfacial mechanical strengths. We connected an outer surface of the bonded sample to a digital spring weight scaler via a solid wire that was firmly attached to the sample surface while using household adhesive glue. Subsequently, we pulled the scaler outward in a direction normal to the sample die until the bonded sample was debonbed, while the weight scaler recorded the maximum force at the point of delamination. For electrical measurements, an Au-Ge-Ni alloy (80:10:10 wt %) and pure Au layers with thicknesses of 30 and 150 nm, respectively, were sequentially C 2020, 6, 28 3 of 10 deposited via thermal evaporation on both of the outer surfaces of the bonded Si pieces, as ohmic electrodes. In this manner, Au/Au-Ge-Ni/Si contacts were formed and covered the entire Si surfaces of the bonded samples. We did not apply any annealing for the contacts to prevent the potential heating effects to the bonded interfacial characteristics. Subsequently, we measured the current-voltage characteristics across the bonded interfaces. For optical measurements of GQDs, a sample containing 3 w/v % of PAM between glass plates was prepared in the same manner as above, albeit with no surface pretreatment. The photoluminescence spectra from the bonded glass samples were then measured with a Si charge-coupled device array spectrometer under the irradiation of a 380-nm-peaked ultraviolet lamp. In this study, for the purpose of simplicity, we used Si wafers as representative semiconductor materials. However, the proposed method can be easily extended to other semiconductors, given that numerous wafer-bonding experimental demonstrations between dissimilar materials are reported to date [4,[26][27][28][29][30][31].

Bonding via Bare GQDs
First, we present the results of the bonding approach: (1) namely, bond formation via directly coating the colloidal GQD solution onto the semiconductor wafers. Figure 1a shows a plane-view scanning electron microscope image of the GQDs on a Si wafer. The samples for observation were lightly coated with Au via sputtering to prevent electric charge-up during electron microscopy. It was observed that the GQDs, with diameters and an areal density of approximately 10-20 nm and 8 × 10 10 cm −2 , respectively, were uniformly dispersed on the Si wafer before bonding. Figure 1b shows a cross-sectional image, which was taken by a regular scanning electron microscope detecting secondary electrons, of the bonded interface, for a sample bonded at 300 • C with HF surface pretreatment. The Si wafers are uniformly and firmly intact, with sufficient endurance during the cleaving process, as shown in the figure. Figure 1c   deposited via thermal evaporation on both of the outer surfaces of the bonded Si pieces, as ohmic electrodes. In this manner, Au/Au-Ge-Ni/Si contacts were formed and covered the entire Si surfaces of the bonded samples. We did not apply any annealing for the contacts to prevent the potential heating effects to the bonded interfacial characteristics. Subsequently, we measured the currentvoltage characteristics across the bonded interfaces. For optical measurements of GQDs, a sample containing 3 w/v % of PAM between glass plates was prepared in the same manner as above, albeit with no surface pretreatment. The photoluminescence spectra from the bonded glass samples were then measured with a Si charge-coupled device array spectrometer under the irradiation of a 380nm-peaked ultraviolet lamp. In this study, for the purpose of simplicity, we used Si wafers as representative semiconductor materials. However, the proposed method can be easily extended to other semiconductors, given that numerous wafer-bonding experimental demonstrations between dissimilar materials are reported to date [4,[26][27][28][29][30][31].

Bonding via Bare GQDs
First, we present the results of the bonding approach: (1) namely, bond formation via directly coating the colloidal GQD solution onto the semiconductor wafers. Figure 1a shows a plane-view scanning electron microscope image of the GQDs on a Si wafer. The samples for observation were lightly coated with Au via sputtering to prevent electric charge-up during electron microscopy. It was observed that the GQDs, with diameters and an areal density of approximately 10-20 nm and 8 × 10 10 cm -2 , respectively, were uniformly dispersed on the Si wafer before bonding. Figure 1b shows a cross-sectional image, which was taken by a regular scanning electron microscope detecting secondary electrons, of the bonded interface, for a sample bonded at 300 °C with HF surface pretreatment. The Si wafers are uniformly and firmly intact, with sufficient endurance during the cleaving process, as shown in the figure. Figure 1c shows a magnified view of the bonded interface of Figure Figure 2 shows typical current-voltage characteristics across the bonded interfaces for various bonding conditions. The ohmic electrical property, which is suitable for device applications, was obtained for the bonding conditions with HF surface pretreatment, as observed in the straight current-voltage curves. We observed that the samples bonded at 300 °C, the central bonding temperature, with and without HF surface pretreatment exhibited interfacial electrical resistivities of approximately 1 and 20 Ω·cm 2 , respectively. This result indicates that the HF surface pretreatment provides significantly higher interfacial electrical conductivity, which can be attributed to the removal of the electrically insulating native oxide layer of SiO2 on the Si wafer surface via the HF  Figure 2 shows typical current-voltage characteristics across the bonded interfaces for various bonding conditions. The ohmic electrical property, which is suitable for device applications, was obtained for the bonding conditions with HF surface pretreatment, as observed in the straight current-voltage curves. We observed that the samples bonded at 300 • C, the central bonding temperature, with and without HF surface pretreatment exhibited interfacial electrical resistivities of approximately 1 and 20 Ω·cm 2 , respectively. This result indicates that the HF surface pretreatment provides significantly higher interfacial electrical conductivity, which can be attributed to the removal C 2020, 6, 28 4 of 10 of the electrically insulating native oxide layer of SiO 2 on the Si wafer surface via the HF solution. The higher interfacial electrical conductivity is preferable for most device applications; and therefore, we focused on the bonding with HF pretreatment. Figure 3 plots the dependence of the interfacial electrical resistivity on the bonding temperature for the samples with HF surface pretreatment. It should be noted that the current-voltage data (such as that in Figure 2) includes all series resistances through the sample. Hence, we independently determined the contact resistance of the metal electrode/semiconductor interfaces while using the transmission line method, and then determined the nominal resistivity at the bonded interface by subtracting it from the slope of the current-voltage curve. A minimum interfacial resistivity is observed in terms of the bonding temperature, as shown in Figure 3. This behavior can be attributed to the trade-off between more stable bond formation and GQD or Si oxidation at higher temperatures. As the highest interfacial electrical conductivity measured, we obtained an interfacial resistivity of 0.26 Ω·cm 2 for the bonding temperature of 300 • C, while the plots in Figure 3 show the average values for each bonding temperature. Such a high interfacial electrical conductance value is considered as being highly preferable for most optoelectronic device applications, such as solar cells [37,38]. solution. The higher interfacial electrical conductivity is preferable for most device applications; and therefore, we focused on the bonding with HF pretreatment. Figure 3 plots the dependence of the interfacial electrical resistivity on the bonding temperature for the samples with HF surface pretreatment. It should be noted that the current-voltage data (such as that in Figure 2) includes all series resistances through the sample. Hence, we independently determined the contact resistance of the metal electrode/semiconductor interfaces while using the transmission line method, and then determined the nominal resistivity at the bonded interface by subtracting it from the slope of the current-voltage curve. A minimum interfacial resistivity is observed in terms of the bonding temperature, as shown in Figure 3. This behavior can be attributed to the trade-off between more stable bond formation and GQD or Si oxidation at higher temperatures. As the highest interfacial electrical conductivity measured, we obtained an interfacial resistivity of 0.26 Ω·cm 2 for the bonding temperature of 300 °C, while the plots in Figure 3 show the average values for each bonding temperature. Such a high interfacial electrical conductance value is considered as being highly preferable for most optoelectronic device applications, such as solar cells [37,38]. solution. The higher interfacial electrical conductivity is preferable for most device applications; and therefore, we focused on the bonding with HF pretreatment. Figure 3 plots the dependence of the interfacial electrical resistivity on the bonding temperature for the samples with HF surface pretreatment. It should be noted that the current-voltage data (such as that in Figure 2) includes all series resistances through the sample. Hence, we independently determined the contact resistance of the metal electrode/semiconductor interfaces while using the transmission line method, and then determined the nominal resistivity at the bonded interface by subtracting it from the slope of the current-voltage curve. A minimum interfacial resistivity is observed in terms of the bonding temperature, as shown in Figure 3. This behavior can be attributed to the trade-off between more stable bond formation and GQD or Si oxidation at higher temperatures. As the highest interfacial electrical conductivity measured, we obtained an interfacial resistivity of 0.26 Ω·cm 2 for the bonding temperature of 300 °C, while the plots in Figure 3 show the average values for each bonding temperature. Such a high interfacial electrical conductance value is considered as being highly preferable for most optoelectronic device applications, such as solar cells [37,38].  Figure 4 shows the dependence of the interfacial mechanical bonding strength on the bonding temperature for the samples with HF surface pretreatment. No bonding was formed or the bonding strengths were lower than our detection limit with respect to the bonded samples at room temperature. Figure 4 clearly indicates that the increase in bonding temperature increases the interfacial mechanical stability. The bonding temperatures at or above 300 • C were observed to provide sufficient interfacial mechanical stability for practical device applications. With respect to the highest data measured, we obtained an interfacial mechanical strength of up to 340 kPa for the bonding temperature of 300 • C, while the plots in Figure 4 show the average values for each bonding temperature. In addition, the optical property of wavelength conversion was observed as severely degraded for the bonded glass samples that were prepared for the optical measurement for approach (1), and this is presumably due to the aggregation of GQDs during the drying out of the solution [39].
C 2020, 6, x FOR PEER REVIEW 5 of 11 Figure 3. Dependence of the interfacial electrical resistivity on bonding temperature for the samples with HF surface pretreatment for bonding approach (1). Figure 4 shows the dependence of the interfacial mechanical bonding strength on the bonding temperature for the samples with HF surface pretreatment. No bonding was formed or the bonding strengths were lower than our detection limit with respect to the bonded samples at room temperature. Figure 4 clearly indicates that the increase in bonding temperature increases the interfacial mechanical stability. The bonding temperatures at or above 300 °C were observed to provide sufficient interfacial mechanical stability for practical device applications. With respect to the highest data measured, we obtained an interfacial mechanical strength of up to 340 kPa for the bonding temperature of 300 °C, while the plots in Figure 4 show the average values for each bonding temperature. In addition, the optical property of wavelength conversion was observed as severely degraded for the bonded glass samples that were prepared for the optical measurement for approach (1), and this is presumably due to the aggregation of GQDs during the drying out of the solution [39].

Bonding via Embedding GQDs in Hydrogel
Second, we present the results for approach (2), wherein we perform bonding by embedding GQDs in a hydrogel material. Figure 5 shows typical current-voltage characteristics across the bonded interfaces for various bonding conditions. In a manner that is similar to the case of the bonding approach (1), the ohmic electrical property was obtained for the bonding conditions with HF surface pretreatment. Figure 6 shows the dependence of the interfacial mechanical bonding strength and electrical resistivity on the PAM concentration for the samples with HF surface pretreatment bonded at room temperature. It is noted that the employment of hydrogel can enable mechanically stable bond formation at room temperature [40][41][42], which was conversely difficult in the case of the bonding approach (1). The hydrogen bonds stemming from PAM presumably causes adhesion to semiconductor surfaces [43,44]. Specifically, hydrogen bonds can form between the -NH 2 groups of PAM and the Si surface terminated by the -OH groups due to the water contained in PAM [43]. In addition, the PAM matrix holds the GQDs and suppresses their sedimentation owing to its viscosity induced by the entanglement of PAM polymer chains. Other hydrogel materials, such as agarose and polyvinyl alcohol, may also be used as bonding agents [42]. The interfacial bonding mechanical strength was observed to gradually decrease when the PAM concentration increases, which is potentially attributed to the spatial non-uniformity of the PAM at the bonded interface due to the increased viscosity of the PAM solution, as shown in Figure 6. As the highest data measured, we obtained an interfacial mechanical strength of up to 350 kPa for the case without the HF pretreatment, a PAM concentration of 3 w/v %, and bonding at room temperature. C 2020, 6, 28 6 of 10 which is potentially attributed to the spatial non-uniformity of the PAM at the bonded interface due to the increased viscosity of the PAM solution, as shown in Figure 6. As the highest data measured, we obtained an interfacial mechanical strength of up to 350 kPa for the case without the HF pretreatment, a PAM concentration of 3 w/v %, and bonding at room temperature. With respect to the interfacial electrical conductivity, in a manner that is similar to the case of the bonding approach (1), the HF surface pretreatment was observed to reduce the resistivity from approximately 10 Ω·cm 2 to 3 Ω·cm 2 on average. Nevertheless, the PAM concentration does not significantly affect the interfacial electrical conductivity, as shown in Figure 6. As the highest interfacial electrical conductivity measured, we obtained an interfacial resistivity of 0.53 Ω·cm 2 for the case with HF pretreatment, a PAM concentration of 1 w/v %, and bonding at room temperature, while the plots in Figure 6 show the average values for each condition. The realization of this level of electrical conductivity is considered to be preferable for many optoelectronic device applications, such as solar cells [37,38]. PAM, a hydrophilic polymer material, can contain water, and the wateroriginated ions act as electrical carriers to induce electrical conductance [44]. Reference 42 discusses the influence of the presence of water in the hydrogel on electronic component fabrication and characteristics.
In contrast to approach (1), no degradation was observed in the optical property, even after the dry out of the samples for bonding approach (2), due to the use of the hydrogel material. This result can be potentially attributed to the effects of the surface passivation and/or aggregation suppression of the GQDs [39,45]. Figure 7 shows the photoluminescence spectra of the bonded glass samples with and without GQDs. The mild twin peaks that were observed around 440 and 490 nm were With respect to the interfacial electrical conductivity, in a manner that is similar to the case of the bonding approach (1), the HF surface pretreatment was observed to reduce the resistivity from approximately 10 Ω·cm 2 to 3 Ω·cm 2 on average. Nevertheless, the PAM concentration does not significantly affect the interfacial electrical conductivity, as shown in Figure 6. As the highest interfacial electrical conductivity measured, we obtained an interfacial resistivity of 0.53 Ω·cm 2 for the case with HF pretreatment, a PAM concentration of 1 w/v %, and bonding at room temperature, while the plots in Figure 6 show the average values for each condition. The realization of this level of electrical conductivity is considered to be preferable for many optoelectronic device applications, such as solar cells [37,38]. PAM, a hydrophilic polymer material, can contain water, and the water-originated ions act as electrical carriers to induce electrical conductance [44]. Reference 42 discusses the influence of the presence of water in the hydrogel on electronic component fabrication and characteristics.
In contrast to approach (1), no degradation was observed in the optical property, even after the dry out of the samples for bonding approach (2), due to the use of the hydrogel material. This result C 2020, 6, 28 7 of 10 can be potentially attributed to the effects of the surface passivation and/or aggregation suppression of the GQDs [39,45]. Figure 7 shows the photoluminescence spectra of the bonded glass samples with and without GQDs. The mild twin peaks that were observed around 440 and 490 nm were presumably due to possible bimodal size distribution of the GQDs. The change in the spectrum of the photon count in Figure 7 verifies that the GQDs embedded in the bonded interface converts the incident light peaking at approximately 380 nm into another bundle of light peaking at approximately 450-500 nm. This type of optical wavelength conversion can be utilized in applications involving solar cells, where the interfacial GQDs absorb ultraviolet light (380 nm) and emit visible light (450-500 nm) to fit the solar spectral irradiance to the crystalline silicon absorption spectral sensitivity. However, the proposed GQD-bonding scheme can be used, irrespective of the material species. Hence, the most suitable combination of semiconductors can be selected on demand for each application. The tunability of the absorption and emission wavelengths via the quantum-dot size is advantageous for the adjustment of each application. Moreover, the use of an ensemble of dots with a size distribution can fit some of the applications, such as photovoltaics, in response to the broad solar spectrum. In this study, we demonstrated photonic down conversion by GQD, but not up conversion. However, with down conversion, the implementation of an efficient GQD-mediated interface in a multijunction solar cell might not be trivial because it would require a relatively large separation in the bandgap energies between the top and bottom subcells, such as that in an (Al) (In)GaN-Si combination. In contrast, up-converting GQDs [32][33][34] could be more beneficial in such an application. For two-dimensional graphene sheets, albeit not quantum dots, we previously developed graphene-mediated semiconductor wafer bonding to form a Si/graphene/Si double heterostructure [46]. Hence, we demonstrated the preparation of two types of graphene-based-material-cored double heterostructures that can provide a basis for high-performance nanophotonic devices in the future. (In)GaN-Si combination. In contrast, up-converting GQDs [32][33][34] could be more beneficial in such an application. For two-dimensional graphene sheets, albeit not quantum dots, we previously developed graphene-mediated semiconductor wafer bonding to form a Si/graphene/Si double heterostructure [46]. Hence, we demonstrated the preparation of two types of graphene-basedmaterial-cored double heterostructures that can provide a basis for high-performance nanophotonic devices in the future.

Conclusions
In the study, we proposed and experimentally demonstrated a semiconductor bonding scheme, which was mediated by GQDs. We investigated two approaches as methods for introducing interfacial GQDs, namely bond formation via directly coating the colloidal GQD solution onto the semiconductor wafers and bonding via GQDs embedded in a hydrogel material. For both cases, we analyzed the dependence of the interfacial mechanical bonding strength and electrical conductivity on bonding conditions, and obtained those sufficient for practical device applications. The use of hydrogel as a matrix agent provided stable bond formation at room temperature. We observed a clear optical wavelength conversion by GQDs at the semiconductor interface, and this can lead to various applications for photon management in integrated devices. The process technique simultaneously allows for bonding formation and interfacial function generation, thereby leading to efficient device production. The proposed bonding scheme can also provide a pathway to fabricate highly efficient double heterostructured optoelectronic devices, which employ GQDs as emitters or absorbers, such

Conclusions
In the study, we proposed and experimentally demonstrated a semiconductor bonding scheme, which was mediated by GQDs. We investigated two approaches as methods for introducing interfacial GQDs, namely bond formation via directly coating the colloidal GQD solution onto the semiconductor wafers and bonding via GQDs embedded in a hydrogel material. For both cases, we analyzed the dependence of the interfacial mechanical bonding strength and electrical conductivity on bonding conditions, and obtained those sufficient for practical device applications. The use of hydrogel as a matrix agent provided stable bond formation at room temperature. We observed a clear optical wavelength conversion by GQDs at the semiconductor interface, and this can lead to various C 2020, 6, 28 8 of 10 applications for photon management in integrated devices. The process technique simultaneously allows for bonding formation and interfacial function generation, thereby leading to efficient device production. The proposed bonding scheme can also provide a pathway to fabricate highly efficient double heterostructured optoelectronic devices, which employ GQDs as emitters or absorbers, such as light-emitting diodes [47][48][49], lasers, optical amplifiers, modulators, photodetectors, and solar cells.
Funding: This research was funded by the Fujikura Foundation, the Kato Foundation for Promotion of Science, the Murata Science Foundation, and the Japan Society for the Promotion of Science (JSPS).