2.1.

Type I and Type II Modifications

For dielectric crystals, waveguide formation depends on the efficient modification of the refractive index of the bulk. There are two major effects of refractive index changes, i.e., types I and II.⁶³ Type I refers to the laser-induced tracks with positive index changes ($\mathrm{\Delta }n>0$) that may be due to the slight structural change of lattices. Such variation may be correlated to the breaking of the symmetry of the irradiated lattices.^83–85 To form type I modification, low-energy pulses are required. Recent investigation reveals that the type I modification may also be due to the phase transition of the laser-irradiated regions.^86,87 Type II refers to laser created filaments in the material, which has negative index changes ($\mathrm{\Delta }n<0$) caused by localized crystalline lattice expansion in the focal volume.⁶³ The type II modification is typically correlated to highly altered lattices in which partial or complete amorphization occurs. For type I modifications, the track region is able to guide light because it has a higher refractive index than the surroundings, whereas for type II, the track has a lower index and cannot be used as a waveguide; however, the expansion of volume in the focal plane gives rise to compression of vicinity, leading to the refractive index increments in the surrounding of the track.⁶⁰ With these typical features, several configurations of waveguides have been developed based on types I and II modifications. Figure 1 shows the typical geometries of four basic laser-written waveguides: single-line, double-line, depressed-cladding, and optical-lattice-like structures. For $\mathrm{\Delta }n>0$ in the track, the single-line can be used to guide light since it naturally constructs a waveguide channel [Fig. 1(a)], and such structures are easy for 3D waveguide fabrication.^88,89 However, in crystals, the single-line structure is usually polarization sensitive^84,85 and does not possess good thermal stability.³² Another disadvantage is that the waveguide core lattice has been distorted by the laser irradiation and, for crystals, the bulk features are destroyed in the waveguide.⁹⁰ Based on type I modification, a multiscan technique can be employed for the fabrication of symmetrical waveguides in which several single-line tracks are overlapped transversal to the writing direction.⁹¹ Such a design is preferred to support waveguiding accessible to long wavelengths with low-loss propagation owing to a significant increase of the index contrast.⁸⁹ For $\mathrm{\Delta }n<0$, the track region has a low index in comparison with the surrounds, and the vicinal zones have a relatively high index that may serve as guiding regions.⁹² In practice, to obtain a nearly symmetric waveguide core, the double-line structure is used: the guiding core is located at the region between two parallel tracks [Fig. 1(b)]. The waveguide core is not located at the tracks that have highly altered lattices. This means that the properties of the bulk crystals are well-preserved in the waveguides^93,94 in which the laser-induced modifications of the original lattices are negligible.⁹⁵ In addition, the formed track has excellent thermal stability, enabling applications under high-power conditions.⁹⁶ There are shortcomings offered by the double-line waveguides, which cannot guide light at longer wavelengths since the dimension of the waveguide core is limited to $<20\mu \mathrm{m}$, and for 3D fabrication it requires a complicated design.⁹⁷ The depressed-cladding structure is a more flexible solution [Fig. 1(c)]. It is usually in the geometry of circles or rectangles,⁹⁸ which consists of a number of type II tracks. Since the diameters of cladding waveguides can be artificially determined, they are capable of guiding long-wavelength light, e.g., MIR regime.^78,80,99 Nevertheless, for 3D waveguides, it also requires a special design. A special configuration is an “optical-lattice-like” structure,¹⁰⁰ which contains a periodically arranged array of tracks [Fig. 1(d)]. Some positions are free of any tracks, e.g., in the center of the structure, which can serve as beam propagation channels. By modulating the geometries of tracks, it is able to achieve beam transformation and shaping.^101,102 Generally, the maximum $\mathrm{\Delta }n$ of type I modification is much lower than that of the type II modification induced by femtosecond laser irradiation.³² For instance, the typical index variation of type I waveguides in ${\mathrm{LiNbO}}_3$ crystal does not go beyond $1\times {10}^{-3}$, whereas this value will increase to (2 to 5) × ${10}^{-3}$ at directly irradiated tracks for type II-based structures.^23,63 For Nd:YAG double-line waveguides, $\mathrm{\Delta }n$ as high as $9\times {10}^{-2}$ at type II filaments has been determined according to $\mu$-Raman and $\mu$-photoluminescence microscopic analysis.⁹² Most recently, quantitative phase microscopy became a popular method to recover the index contrast of the waveguides from the phase image. In particular, the index change of $2.5\times {10}^{-3}$ has been maintained after heating up to 1400°C for a depressed cladding waveguide inscribed in a sapphire crystal that exhibited $4.75\times {10}^{-3}$ prior to the annealing process.⁸⁰ Table 1 summarizes the advantages and disadvantages of different configurations in transparent solid materials (crystal, glass, polymer, ceramic, etc).

Fig. 1

Basic geometries of laser writing of waveguides: (a) single-line, (b) double-line, (c) depressed-cladding, and (d) optical-lattice-like configurations. The dark (gray) regions represent the laser-induced tracks.

Table 1

Advantages and disadvantages of different configurations in transparent material.

One could realize the evolution of laser-written tracks from type I to type II via manipulating the laser inscription parameters. For instance, in ${\mathrm{LiTaO}}_3$ crystal, with the low-repetition rate of femtosecond laser pulses (25 kHz), the tracks can be changed from type I to type II. Figure 2(a) shows the influence of incident pulse energy on the geometry of tracks at a fixed laser scanning speed. With the increase of pulse energy, the type I region gradually moves down along the direction that the laser propagates and an elongated filamentation with type II modification appears. This can be explained by the competition of self-focusing and plasma defocusing when the peak power is higher than the critical power $P_{\mathrm{cr}}$:³²

Fig. 2

Optical microscope images of laser-induced tracks with (a) different pulse energies and (b) scanning velocity in ${\mathrm{LiTaO}}_3$ crystal. (c) Different refractive index change profiles from positive to negative, depending on the propagation wavelength in ZnSe crystal. Image (c) is reproduced with permission from Ref. 103, Creative Commons Attribution License (CC-BY).

2.2.

Control of Laser-Induced Tracks

The track morphology is crucial for laser inscribed waveguides because the track geometry plays an important role in the refractive index changes and stress fields. In addition, track engineering is also essential for nonlinear domain modification and well-defined microstructures.

2.2.1.

Effect of writing parameters

For a given laser writing system, the main parameter that determines the localized modification in focal volume is the pulse intensity (or fluence).³² At low pulse intensity but beyond the optical modification threshold, an isotropic refractive index change is achieved,¹⁰⁵ whereas a birefringent refractive index change has been obtained at intermediate intensity.¹⁰⁶ Higher pulse intensity will lead to empty voids caused by microexplosions,¹⁰⁷ depending on the properties of the material. The transverse intensity distribution of the laser beam could be written as¹⁰⁸

Eq. (2)

$$I(r,z)={\left(\frac{w_0}{w(z)}\right)}^2\exp (-\frac{2r^2}{w^2(z)}),$$

where $r$ is the radial distance and $z$ is the axial distance from the beam waist. $w(z)$ is the variation of the laser beam spot size (the distance from the beam axis where the intensity drops to $1/e^2$ of the maximum value) and is given by

Eq. (3)

$$w(z)=w_0\sqrt{1+{\left(\frac{z}{z_0}\right)}^2},$$

and $w_0$ is the diffraction-limited minimum beam waist radius:

Eq. (4)

$$w_0=\frac{M^2\lambda }{\pi \mathrm{NA}},$$

where $M^2$ is the Gaussian beam quality factor and NA is the numerical aperture of the focusing objective. The Rayleigh range of the focus within a transparent material having a refractive index $n$ is given by

Eq. (5)

$$z_0=\frac{M^{2}n\lambda }{\pi {\mathrm{NA}}^2}.$$

The laser beam focus size at the focal plane is simply given by the $w_0$ and $z_0$ in the transverse and axial directions of the incident beam when neglecting optical aberrations and nonlinear effects. The peak intensities reached at the focal volume of the transparent material are:

Eq. (6)

$$I_p=\frac{E_p}{\pi w_0^2{\tau }_p},$$

where $E_p$ is the pulse energy and ${\tau }_p$ is the pulse duration. As we can see, the beam waist determines the transversal size of the modifications induced in the dielectrics; together with the scanning velocity, they define the pulses superposition and the total amount of the energy delivered at a focal spot. It also defines the pulses that have enough intensity to promote changes in the material.

Generally, the higher the energy of laser pulses is, the more modified the tracks are.⁹⁰ In addition, higher-speed scanning creates weaker tracks.²³ This is easily understandable because the track geometry is determined by the average energy deposition of the unit volume in the samples. Figure 3(a) shows the laser-induced tracks in a $\mathrm{Nd}:{\mathrm{YVO}}_4$ crystal at different energies and scanning rates (the repetition rate of the laser is set at 1 kHz). Similar effects are observed for other crystals. Therefore, the choice of energy and scanning rate for a successful waveguide fabrication is to balance both effects on dielectric materials. The energy of pulses is typical of micro-Joule ($\mu \mathrm{J}$) or sub-$\mu \mathrm{J}$, and the used scanning rate ranges from tens of $\mu \mathrm{m}/\mathrm{s}$ to tens of mm/s. The polarization of femtosecond laser pulses may also affect the track writing, although in most cases, the effect is negligible. For example, in an optical-lattice-like structure of Nd:YAP crystal,⁷⁵ as femtosecond laser polarization is perpendicular to the scan direction, the waveguide cores are located in the regions between adjacent tracks, while in the center of the structure in the case of polarization parallel to the scan orientation. Figure 3(b) shows the optical microscope image of the lattice-like structure in a Nd:YAP crystal and the guided modes obtained under different laser polarizations.

Fig. 3

(a) Microscope images of tracks with different pulse energies and scanning rates in $\mathrm{Nd}:{\mathrm{YVO}}_4$. (b) Influence of different laser polarizations on the tracks and modal profiles in Nd:YAP, perpendicular to the scans (No. 1) and parallel to the scans (No. 2), respectively. Images reprinted from Ref. 75, © 2016 The Optical Society (OSA). (c) Optical images of tracks as the laser writing along different crystalline orientations in $\mathrm{Nd}:{\mathrm{GdVO}}_4$. (d) Laser-induced multifoci in ${\mathrm{LiTaO}}_3$, corresponding horizontal waveguides, and modal profiles. Reprinted with permission from Ref. 109, © 2019 IEEE.

2.2.3.

Beam shaping techniques

During the process of femtosecond laser irradiation, the laser beam is generally fixed and focused at a certain depth beneath the sample surface. The sample is usually mounted on a computer-controlled $XYZ$ linear stage to realize an arbitrary shape, and a CCD system is used to monitor the waveguide writing process. The sample can be translated along the beam propagation direction, i.e., longitudinal writing, or orthogonally to, i.e., transverse writing.⁶⁰ The waveguide formed by longitudinal writing has intrinsic circular symmetry, while it is restricted by the working distance of the objective lens and it has limitations in 3D layouts.¹¹⁴ In contrast, the design flexibility of transverse writing is enormous and allows for the fabrication of 3D structures,^23,60 in spite of their asymmetric cross sections.

Over the past decades, beam-engineering techniques have been developed to improve the shape of the laser-written waveguides, such as the double-beam technique,¹¹⁵ adding cylindrical lenses,¹¹⁶ the slit technique,⁴¹ using a nondiffractive Bessel beam,¹¹⁷ and using a spatial light modulator (SLM).³⁹ Figure 4(a) illustrates the writing process of depressed cladding waveguides utilizing an ellipsoidal focal spot, where a number of tracks are inscribed one by one, thus, resulting in a time-consuming process. Nowadays, SLMs have been used widely in femtosecond laser processing. From the work of Qi et al., the fabrication times have been considerably reduced,⁴¹ as shown in Fig. 4(b). The phase mask for writing horizontal lines is presented in Fig. 4(c). Thereby, horizontal tracks can be employed to construct the cladding of waveguides in a ${\mathrm{LiNbO}}_3$ crystal that exhibit the polarization-independent guidance⁴² [Figs. 4(d)–4(f)]. Later in 2019, Zhang et al.¹¹⁸ proposed a single-scan method to fabricate cladding waveguides using 3D engineered focal field [Figs. 4(g)–4(j)]. The longitudinally oriented ring-shaped focal field is composed of 16 discrete spots, allowing for the generation of annular cladding waveguides at a single shot of laser writing, demonstrated in Figs. 4(k)–4(m). Most recently, Sun et al.³⁹ developed a new and efficient multifoci-shaped femtosecond pulsed method to write circular symmetric waveguides with an SLM.

Fig. 4

Schematic diagram of direct laser-written cladding waveguides (a) with an ellipsoidal focal spot and (b) with a slit-shaped beam focus. Images (a), (b), and (g) are reprinted with permission from Ref. 40, © 2017 OSA. (c) The phase mask for writing horizontal lines. Microscope image of (d) horizontal tracks, (e) waveguide, and (f) near-field profiles. Images (c)–(f) are reproduced with permission from Ref. 42, CC-BY. (g) Schematic plot of single-scan cladding waveguides utilizing a longitudinal ring-shaped focal field. (h) Calculated 3D isosurface, (i) phase mask, and (j) simulated focal intensity profile. (k) Microscope image and (l), (m) corresponding modal profiles. Images (h)–(m) are reprinted with permission from Ref. 118, © 2019 Chinese Laser Press (CLP).

2.2.4.

Nonlinear domain engineering

The morphology of tracks is crucial for not only waveguide configuration but also for engineering the crystal properties to achieve modulation of nonlinearity polarization. Especially for type II modification, focused laser pulses will cause highly damaged lattices, where an amorphous volume forms that reduces the nonlinear coefficient. Such a modification can be periodically introduced to realize spatial modulation of the second-order nonlinear coefficient ${\chi }^{(2)}$, enabling quasi-phase-matching (QPM) structures,¹¹⁹ also known as nonlinear photonic crystals (NPC).¹²⁰ Thereby, the researchers have fabricated 3D NPCs by selectively erasing the nonlinear coefficients of a ${\mathrm{LiNbO}}_3$ crystal.¹²¹ The recently developed 3D NPCs present efficient generations of second-harmonic vortex and Hermite–Gaussian beams, showing the potential for nonlinear beam-shaping devices⁶ and nonlinear volume holography^122,123 in comparison to the two-dimensional (2D) case. Additionally, the huge flexibility of direct laser writing enables complex 3D domain structures,^43,120 allowing their monolithic integration with other phase modulators and switches, and improving the performances and functions of on-chip devices. Using the same strategy of domain erasing in a quartz crystal, researchers from Shandong University have realized tunable second harmonic generation (SHG) down to the ultraviolet wavelength.¹²⁴ In addition, there will be an obvious thermal effect at the high-repetition rate of femtosecond laser pulses, which can be utilized to realize reversed polarization. The group from Australia reported on the similar 3D domain structures in a BCT crystal by femtosecond laser direct writing through the thermoelectric-field poling,¹²⁵ allowing for precise domain inversion with high resolution. This laser engineering strategy can be extended to other systems of crystals.

Moreover, laser-induced quasi-phase-matching (LiQPM) grating structures inside the waveguide’s core allow efficient frequency conversion, which opens new avenues for advanced all-integrated nonlinear devices. For instance, using femtosecond laser-induced domain inversion in ${\mathrm{LiNbO}}_3$ waveguides, a conversion efficiency of 17.45% for frequency doubling of 815 nm has been demonstrated.¹²⁶ In addition, a depressed cladding waveguide-integrated LiQPM grating scheme has been proposed by Denz et al.¹²⁷ With the precise and flexible 3D domain engineering, as shown in Figs. 5(a) and 5(b), they have realized broadband and multiwavelength frequency conversions.⁴³ Particularly, the second harmonic vortex can be demonstrated by a helical twisted domain grating into a waveguide, as depicted in Figs. 5(c)–5(e).

Fig. 5

(a) Schematic design of a waveguide-integrated LiQPM grating; (b) SH microscope image of LiQPM grating and waveguide; and (c) helical grating structure. Microscope image of the helical structure: (d) the front face and (e) top view. Images reprinted with permission from Ref. 43, © 2020 OSA.

2.2.5.

Well-defined micro- and nanostructures

The laser-written track is also promising for creating spatially well-defined microstructures in combination with chemical wet etching. In the laser-modified volumes, the femtosecond pulses have driven the crystal network to a predamaged state that shows a larger selective etching than the unperturbed bulk lattices.¹²⁸ A good sample is the YAG crystal, as shown in Fig. 6(a), where the optical-lattice-like waveguide is fabricated by direct laser writing, and then the sample is immersed into the selected acid (e.g., phosphoric acid).¹²⁹ As one can see, after some time, the etched portion of the damaged track is filled with air, thus resulting in a larger refractive index change than the unetched region [Figs. 6(b) and 6(c)]. Therefore, the confinement of the light field will be enhanced. In this work, the etched microstructured optical waveguide supports good guidance at the MIR wavelength of $4\mu \mathrm{m}$ [Figs. 6(d) and 6(e)], which is impossible for structure before etching. Actually, early in 2013, Kar et al. produced large microfluidic channels within YAG by selectively wet etching laser-written type I tracks.¹³⁰ The technique is very interesting and offers an efficient solution to 3D nanolithography of crystals. In 2019, Osellame et al.¹³¹ reported on the nanopores with mm-scale length. By implementing auxiliary vertical etching pores to build a 3D-connecting etching architecture, they have achieved 3D nanolithography with a long length. Using this strategy, they have produced subwavelength diffraction gratings and microstructured waveguides in YAG crystals.

Fig. 6

(a) Microscope images of hollow optical-lattice-like structures at different etching times in YAG crystal. (b) Before polished and (c) after polished. Near-field modal profiles at $4\mu \mathrm{m}$ along (d) TM and (e) TE polarization, respectively. Images reproduced with permission from Ref. 129, © 2020 CLP.

2.3.

Mode Guidance Correlated to Crystal Systems

Dielectric crystals are important media for optical and photonic applications, which can be roughly cast into a few groups, such as laser crystals, nonlinear optical crystals, and electrooptic crystals.²³ Unlike the amorphous substrates, crystalline material entails structures of long-range orders of lattices, offering unique physical properties that are not exhibited in glass, such as even order optical nonlinearities and birefringence.¹ There are seven crystal systems to characterize the crystal lattices or structures of different geometries, i.e., cubic, tetragonal, hexagonal, trigonal, orthorhombic, monoclinic, and triclinic. Up to the current date, a wide range of crystals has been used for the fabrication of direct laser-written waveguides; for example, YAG crystal belongs to the cubic, ${\mathrm{LiNbO}}_3$, $\mathrm{\beta }\text{-}\mathrm{BBO}$, and sapphire crystals belong to the trigonal, and 6H-SiC crystal belongs to the hexagonal crystal family. To some extent, the symmetry of each lattice determines the waveguide properties, such as polarization dependence and propagation loss. Table 2 summarizes the waveguide configurations and guiding properties of typical crystals in different crystal systems. For each lattice family, a brief overview will be given below.

Table 2

Summary of latest published works about waveguide configuration and properties of typical crystals in different crystal systems.

For cubic systems, YAG is one of the most well-studied crystals in the direct laser writing of waveguides. For the depressed-cladding and optical-lattice-like waveguides, symmetric guidance for any polarization, i.e., both transverse magnetic (TM) and electric (TE) are supported,^100,132 whereas the dual-line geometry only allows TM-polarized guidance,⁷⁹ which is mainly related to the high symmetry of the cubic system that possesses three fourfold axes of rotation. The cubic-lattice crystals exposed to laser irradiation undergo a severe modification that results in a positive index change in a specific polarization. A similar polarization-dependent guiding behavior is also found in Nd:GGG crystal,¹³⁴ another typical cubic crystal. Although for the tetragonal system, such as $\mathrm{Nd}:{\mathrm{YVO}}_4$ and $\mathrm{Nd}:{\mathrm{GdVO}}_4$, only one fourfold axis of rotation is contained, leading to a lower symmetry than that of the cubic system. As a result, the dual-line structure supports guidance along both TE and TM polarizations^138,141 but is still of slightly anisotropic guidance. The depressed-cladding waveguides in the tetragonal is guided in an anisotropic way as well, which is different from the cubic YAG crystal. As for the hexagonal system, up to now, only 6H-SiC crystal is employed for laser writing of waveguides.¹⁴³ Like the cubic system, both TM and TE guidances are determined in rectangular cladding waveguides, whereas only TM guiding mode is supported in dual-line structure. In hexagonal system, there is one six-fold axis of rotation, thus the symmetry is between the cubic and tetragonal systems.

${\mathrm{LiNbO}}_3$, ${\mathrm{LiTaO}}_3$, $\mathrm{\beta }\text{-}\mathrm{BBO}$, and sapphire, as the typical crystals belonging to the trigonal system, exhibit unique nonlinearities and birefringence. Therefore, as these crystals are processed under laser writing, there will be some different structural modifications due to the high-order threefold axis of rotation. Various configurations have been inscribed in ${\mathrm{LiNbO}}_3$ crystal by femtosecond lasers, such as type I modified single-line and multiline structures, type II modified dual-line waveguides, depressed-cladding, optical-lattice-like, and ride waveguides. Generally, in ${\mathrm{LiNbO}}_3$ and ${\mathrm{LiTaO}}_3$ crystals, type I and type II waveguides would suffer from serious polarization-dependent guiding, which only permits polarization along the orientation parallel to the laser-induced tracks,^91,147 whereas the depressed-cladding waveguide enables isotropic guiding as well as in Ti-sapphire crystal.¹⁶⁵ However, for $\mathrm{\beta }\text{-}\mathrm{BBO}$ crystal, only TM-polarized guiding is supported in depressed-cladding waveguides.¹⁵⁰ In addition, polarization-dependent guiding also happens in both optical-lattice-like¹⁴⁵ and ridge configurations¹⁴⁶ in a ${\mathrm{LiNbO}}_3$ crystal, which only allows $n_0$-polarized guidance.

In addition to the crystal system mentioned above there are lower-symmetry crystal families that do not contain a higher-fold axis of rotation, such as orthorhombic system and monoclinic system. In such crystal systems, laser-written waveguides present isotropic guiding behavior. For instance, ${\mathrm{KTiOPO}}_4$ (KTP) is a typical crystal belonging to the orthorhombic system; the guidance for dual-line,¹⁵⁵ depressed-cladding,¹⁵⁶ and optical-lattice-like waveguides¹⁵⁷ is almost polarization independent. In addition, ${\mathrm{BiB}}_3{\mathrm{O}}_6$,¹⁶⁰ Nd:KGW,¹⁶⁴ and $\mathrm{KLu}({\mathrm{WO}}_4)$¹⁶⁶ crystals belong to the monoclinic system, another member of the lower-symmetry family. In these crystals, femtosecond laser-written waveguides mainly focus on the depressed-cladding configuration, which allows symmetric guidance for both TM and TE polarizations due to the poor lattice symmetry. These features imply the unique modification of laser pulses on the crystals with different symmetries.

2.4.

Mode Modulation

In recent years, numerous newly developed waveguide geometries that allow mode modulation and guiding polarization engineering have been demonstrated, benefiting from the rapid and direct prototyping of femtosecond laser writing.

For type I modification, the multiscan technique offers a flexible manner to tailor the mode profiles by controlling the number and separation of laser-induced tracks⁹¹ [Fig. 7(a)]. In addition, by arranging the laser-modified tracks into an annular ring, the ring-shaped mode profiles in BGO crystal have also been obtained,¹³⁶ as shown in Fig. 7(b). In particular, such a design allows for polarization-independent guiding, which is different from single- and multiline type I waveguides that only support one special polarized guiding. In addition, the rings can be inscribed in the desired diameter to meet the practical applications. Theoretically, one could realize mode modulation with arbitrary geometries by proper design and arrangement of the laser-induced tracks.

Fig. 7

(a) Microscope images and modal profiles of tailored multiline waveguides in a ${\mathrm{LiNbO}}_3$ crystal, reproduced with permission from Ref. 91, © 2018 Elsevier. (b) Ring-shaped waveguide based on type-I modification in a BGO crystal, reprinted with permission from Ref. 136, © 2017 OSA. (c) Polarization engineering for dual-line waveguides in a ${\mathrm{LiNbO}}_3$ crystal, reproduced with permission from Ref. 167, © 2020 Elsevier. (d) The “ear-like” waveguide in Nd:YAG crystal, reprinted with permission from Ref. 168, © 2021 OSA. (e) Double-cladding waveguide in $\mathrm{Nd}:{\mathrm{YVO}}_4$ crystal, reprinted with permission from Ref. 169, © 2019 OSA.

As for type II modification, one can fabricate a series of dotted lines along the horizontal direction via multiscan, modulating the modal profiles along the TE polarization owing to the changed stress field distribution,¹⁶⁷ as shown in Fig. 7(c). At recent reports in SiC crystal, the mode modulation of waveguides has also been achieved by controlling the morphology of laser-induced tracks.¹⁴³ In addition, femtosecond laser-written waveguide arrays based on type II modification can implement mode modulation via the effect of evanescent coupling. For example, the one-dimensional planar waveguide laser arrays in Nd:YAG crystals support flexible discrete diffraction by adjusting the coupling condition.¹⁷⁰ Zhang et al.³⁷ reported the 2D waveguide arrays in ${\mathrm{LiNbO}}_3$ crystal, where the output mode profiles can be modulated as the pump beam goes into the array at different positions.

Beyond the above reports, some designs based on the depressed-cladding structures also enable mode modulation. Sun et al. proposed an “ear-like” waveguide structure with extra tracks at the sides of cladding, which is an evolved scheme of double-line and depressed-cladding waveguides that was first reported by Okhrimchuk et al.¹⁷¹ Such a design is intended to further enhance the optical confinement of the light field. Therefore, more symmetric mode profiles and lower propagation losses are preferred compared with the normal cladding waveguides,¹⁶⁸ as shown in Fig. 7(d). Another interesting sample is the double-cladding waveguide [Fig. 7(e)], which is composed of two concentric tubular shapes with different diameters.¹⁷² One of the advantages is that the outer cladding could enhance the coupling efficiency of the pump beam and inner cladding permits an output beam with low-order modes, i.e., generating an integrated single-mode laser system.¹⁶⁹ Moreover, using a third-cladding structure, Wu et al.¹³² demonstrated annular ring waveguide lasers, promising for complex integrated pump sources to alleviate thermal lensing.

3.1.

Y-Branches

As the basic 3D components, a $\mathrm{Y}$ branch-based splitter allows for dividing the single-input beam into several output beams in different dimensions. For example, the 3D waveguide beam splitters ($1\times 4$) based on type-I modification have been realized in ${\mathrm{LiNbO}}_3$⁸⁸ and BGO crystals,⁸⁹ whereas the polarization-dependent properties are very different for the two crystalline waveguides. The ${\mathrm{LiNbO}}_3$ only supports light propagation along particular orientations, whereas wave guidance in BGO is isotropic for any transverse polarization. As for type-II modification, $\mathrm{Y}$-branches work well in a relatively simple way for 2D structures, as reported in diamond¹⁷⁸ and YAG crystals.⁹⁷ However, in the 3D case, disadvantages are emerging due to the unbalanced guiding properties and additional architecture required to connect separate channels. Moreover, the geometry of depressed cladding supports wave-guiding along any polarization, and for 3D fabrication it also requires a complicated design. The buried $\mathrm{Y}$-junctions based on circular and rectangular depressed-cladding geometries have been proposed in YAG^179,180 and Ti:sapphire crystals,¹⁸¹ as shown in Figs. 8(a)–8(e), which produce smooth transitions and preserve modal profiles. Furthermore, Ajates et al. reported the novel 3D beam splitting based on depressed-cladding architecture in ${\mathrm{LiNbO}}_3$ crystals [Fig. 8(f)], realizing the desired 3D guiding path.¹⁸² Although the TM-polarized modes have higher losses than that of TE, the 3D structures work well in the near-infrared band. It is possible to implement any arbitrary 3D photonic circuit in crystalline materials based on circular cladding waveguides.

Fig. 8

(a) Fabrication and 3D schematic diagram of $\mathrm{Y}$-splitters based on rectangular cladding geometry in Ti:sapphire crystal, (b) microscope image of 1-deg branching angle, and (c) intensity distributions at 1064 nm. Images (a)–(c) are reproduced with permission from Ref. 181, © 2018 Elsevier. Microscope images of $\mathrm{Y}$-branch with circular cladding structure (d) in top view and (e) in cross section, as well as modal profiles of two arms. Images (d) and (e) are reproduced with permission from Ref. 179, © 2017 Elsevier. (f) 3D beam-splitting structures in a ${\mathrm{LiNbO}}_3$ crystal, reprinted with permission from Ref. 182, © 2018 Optica.

3.2.

Photonic-Lattice-Like Structures for 3D Microfabrication

Additional branched boundaries are essential for beam splitters in Sec. 3.1; as a result, the fabrication processing is usually time-consuming. To this end, the family for 3D microfabrication is designed, known as an “optical-lattice-like” photonic structure, which contains a number of periodic tracks with geometry defects. A typical structure has a track layer of hexagonal layout, where the light field is confined at enclosed track-free regions. Particularly, better guiding confinement is followed at multilayers rather than single layer. Tailored beam steering in the 3D manner, such as beam splitting or ring-shaped transformation,¹⁰⁰ can be realized by directly introducing additional defect tracks at certain positions. These defect lines served as beam blocks or confiners that smoothly reshape the guiding profiles during the propagation; therefore, no additional losses are introduced.¹⁰² Based on this strategy, 3D beam steering of ring-shaped transformation has been successfully inscribed in YAG crystals,¹⁰⁰ as shown in Figs. 9(a) and 9(b). In addition, highly compact 3D beam splitters ($1\times N$) have been implemented in many platforms, such as ${\mathrm{LiNbO}}_3$,¹⁴⁵ ${\mathrm{KTiOPO}}_4$,¹⁵⁷ Nd:YAG, and Ti-sapphire crystals¹⁵³ [Figs. 9(c)–9(e)]. By carefully considering the crystalline properties, more advanced applications, such as nonlinear frequency conversion¹⁸³ and miniature laser sources have been demonstrated.¹⁰¹ Such 3D designs could efficiently improve the integration of complex photonic circuits. Considering the huge differences in the characteristics of crystals, however, each type requires a different 3D inscription and feasible optimization scheme.

Fig. 9

(a) Schematic illustration of $1\times 4$ beam splitting and ring-shaped transformation based on photonic-lattice-like structures. Image (a) is reproduced from Ref. 102. (b) Measured evolution of ring-shaped transformation in a Nd:YAG crystal. The scale bar is $50\mu \mathrm{m}$. (c) Prototype design and microscope images of $1\times 3$ beam splitters in a ${\mathrm{LiNbO}}_3$ crystal, (d) measured and (e) simulated modal profile. Images (c)–(e) are reprinted with permission from Ref. 145, © 2016 IEEE.

3.3.

3D Polarizers

In photonic integrated circuits (PICs), the undesirable effect of polarization dependence limits the practical applications of polarization-division multiplexed systems. Hence, polarization diversity devices including polarizers, polarization beam splitters (PBSs), and polarization rotators attract researchers’ interest. Femtosecond laser writing provides an on-demand fabrication method for polarization photonic devices. Recently, a new structure of polarizer has been employed to engineer the guiding polarization in ${\mathrm{LiNbO}}_3$ crystals,¹⁸⁴ as depicted in Figs. 10(a) and 10(b), which is composed of three parts: a rectangular-shaped channel, a $1\times 2$ beam splitter acting as the connection element, and two output channels. In such a structure, the output will be well-separated into TE and TM polarized light as the polarized light or hybrid light of TE plus TM modes enters the waveguide [Fig. 10(c)]. The polarization extinction ratio and insertion loss can be comparable with that of fused-silica-based PBS. In addition, a 3D polarizer in a cross-shaped geometry has been produced in ${\mathrm{LiNbO}}_3$ crystal,¹⁶⁷ consisting of a square cladding waveguide and four vertically oriented double lines [Figs. 10(d) and 10(e)]. This 3D configuration enables specific polarized-light output at different waveguide arrays, i.e., the TM or TE-polarized light can be controllably coupled into vertical or horizontal waveguide arrays, respectively.

Fig. 10

(a) Microscope images in top view, (b) end-face of polarization beam splitters, and (c) modal profiles along $n_0$, $n_e$, and circular polarizations, respectively. Images (a)–(c) are reprinted with permission from Ref. 184, © 2020 IEEE. (d) Schematic plot and microscope images of 3D polarizer. (c) Modal profiles at different polarizations. Images (d) and (e) are reprinted with permission from Ref. 167, © 2020 IEEE.

3.4.

Other Schemes

Tapered waveguide structures, like the tapered optical fibers, have different circular radii of input and output arms. In 2020, Romero et al.¹⁸⁵ proposed a 3D tapered waveguide by smoothly decreasing the separation between the tracks that consisted of the depressed-cladding waveguide, as shown in Fig. 11(a). This approach enables the desired reduction of the input/output radius, allowing highly multimodal coupling into single-mode waveguides, as depicted in Fig. 11(b). It is particularly interesting for waveguide laser systems, which reaches a higher coupling efficiency of the pump beam and enhanced intensity of the output power compared to a straight waveguide with the same radius. In addition, such a 3D design offers an efficient manner for modal formatting or light concentration [Fig. 11(c)]. Another promising scheme is the 3D waveguide array,³⁷ which has a potential application in 3D photonic devices, such as photonic lanterns. Figure 11(d) illustrates the prototype of 3D waveguide arrays in Nd:YAG crystals,¹⁸⁶ which are composed of seven adjacent hexagonal cladding structures at the input face and have each waveguide separated radially from the center at a certain distance at the output face, as shown in Fig. 11(e). This design shows a good wave guiding for both the visible and near-infrared bands. The directional coupler is the basic building block for integrated classical and quantum photonic circuits.⁴⁵ Figure 12(a) presents the 3D design of a $3\times 3$ directional coupler based on a depressed-cladding waveguide in a ${\mathrm{Tm}}^{3+}:\mathrm{YAG}$ crystal¹⁸⁷ in which several cladding tracks are removed to form the coupling region [Fig. 12(b)]. The top and cross-section view microscope images are shown in Figs. 12(c) and 12(d). Single-mode guidance with an output splitting ratio of 19:40:41 at 810 nm was demonstrated in this work [Fig. 12(e)], indicating the potential applications in quantum storage devices. By integrating embedded 3D electrodes on both sides of the coupling region of the buried directional coupler in a ${\mathrm{LiNbO}}_3$ crystal,¹¹⁸ as shown in Fig. 12(f), the efficiency of electro-optic modulation was finely improved [Fig. 12(g)].

Fig. 11

(a) Microscopic pictures of a tapered cladding waveguide in a Nd:YAG crystal and (b) modal profiles at the input radii of $24\mu \mathrm{m}$ and output of $6\mu \mathrm{m}$, respectively. (c) Modal profiles of a straight and tapered waveguide at the same output radii from an incident LED light, reproduced with permission from Ref. 185, CC-BY. (d) Prototype of depressed-cladding 3D waveguide arrays. (e) Optical micrographs at the output face for different separations between the central and adjacent waveguides. Images (d) and (e) are reprinted with permission from Ref. 186, © 2017 IEEE.

Fig. 12

(a), (b) Geometry and cross-section design in the interaction region of the $3\times 3$ directional coupler in a ${\mathrm{Tm}}^{3+}:\mathrm{YAG}$ crystal. (c), (d) Top and output view microscope images and (e) output intensity distribution. Images (a)–(e) are reproduced with permission from Ref. 187, CC-BY. (f) Schematic plot of $2\times 2$ directional coupler integrated with 3D microelectrodes in a ${\mathrm{LiNbO}}_3$ crystal. (g) Output intensity profiles with different voltages. Images (f) and (g) are reprinted with permission from Ref. 118, © 2019 CLP.

4.1.

Waveguide Lasers

Benefitting from the high optical confinement, femtosecond laser-written waveguides provide efficient miniaturized laser sources that cover the spectral range from visible to MIR (530 nm to $4\mu \mathrm{m}$) with many advantages, such as reduced lasing thresholds, enhanced slope efficiencies, compact size, nondiffraction beams, and multifunctional integration.⁷ Up to now, some notable improvements have been achieved, depending on the active ions and the mechanisms for lasing. Table 3 summarizes the reported results for waveguide lasers emitting at different wavelengths based on various cavity designs fabricated in femtosecond laser writing.

Table 3

Summary of reported results for waveguide lasers emitting at different wavelengths based on various laser-cavity designs.

Laser emission at visible wavelengths mainly relies on ${\mathrm{Pr}}^{3+}$-doped gain media, such as $\mathrm{Pr}:{\mathrm{SrAl}}_{12}{\mathrm{O}}_{19}$, $\mathrm{Pr},\mathrm{Mg}:{\mathrm{SrAl}}_{12}{\mathrm{O}}_{19}$, and $\mathrm{Pr}:{\mathrm{LiYF}}_4$ crystals. For instance, using well-established dual-line structure, a green waveguide laser operating at 525.3 nm was first demonstrated in $\mathrm{Pr},\mathrm{Mg}:{\mathrm{SrAl}}_{12}{\mathrm{O}}_{19}$ crystals.¹⁹⁰ Since then, based on the rhombic cladding geometry, orange and deep red laser oscillations at 604 and 720 nm were produced in $\mathrm{Pr}:{\mathrm{LiYF}}_4$ crystal.¹⁹¹ Moreover, laser-written active dual-line waveguides in Ti:sapphire have offered tunable laser operating from 700 to 870 nm.¹⁹² In addition, with the same dual-line configuration in Ti:sapphire, Grivas et al.¹⁵² reported a femtosecond continuous-wave mode-locked (CWML) laser emitting at 790 nm. The generated waveguide laser was operated at pulses of 41.4 fs with a fundamental repetition rate up to 21.25 GHz.

Near-infrared spectral range is the most common laser emission band, which is gained by writing an active waveguide into well-known ${\mathrm{Nd}}^{3+}$- and ${\mathrm{Yb}}^{3+}$-doped crystals. Beyond a normal straight laser cavity, it is also possible to achieve efficient optical feedback for lasing oscillation in curved and branched cavity geometry owing to an enhanced optical gain of a waveguide laser. For example, in femtosecond laser-written 3D structures, compact waveguide lasers at $1\mu \mathrm{m}$ have been proposed in different functionalities, such as $\mathrm{S}$-curved,^159,194 $\mathrm{Y}$-branches,^{97,180,208,216} $1\times 2$ and $1\times 4$ beam splitting,¹⁰¹ ring-shaped beam transformation,^100,132 and optical-lattice-like lasing⁷⁵ [Figs. 13(a)–13(e)]. Laser performance in these reports is comparable to that of straight waveguides, suggesting a superior flexibility for prototyping laser sources using femtosecond laser writing. By further integrating a suitable absorber, passively $Q$-switched lasing has been realized. For example, using a high gain Yb:YAG channel waveguide, an efficient $Q$-switched laser has shown a slope efficiency of 74% with an average power up to 5.6 W in a single-pass optical pumping without any reflective mirrors.¹⁹⁵ Most recently, mode-locked waveguide lasers operating at multi-gigahertz (GHz) regimes have attracted increasing interest. For instance, 6.5 GHz $Q$-switched mode-locked (QML) lasers have been demonstrated in a femtosecond laser-written $\mathrm{Nd}:{\mathrm{YVO}}_4$ cladding waveguide using 2D materials as the saturable absorber.²⁰² Another recent sample is Nd:YAP crystal : 31.6-GHz QML lasers have been demonstrated in an $\mathrm{S}$-curved waveguide,¹⁵⁹ as shown in Fig. 13(f). In particular, both single- and dual-wavelength (1064 and 1079 nm) lasing can be achieved by tuning the pump polarization [Figs. 13(g) and 13(h)].

Fig. 13

Complex waveguide laser modal profiles at $1\mu \mathrm{m}$: (a), (b) $\mathrm{Y}$-branches, (c) $1\times 4$-branch, (d) ring-shaped transformation, and (e) optical-lattice-like. Images (a) and (d) are reproduced with permission from Ref. 100. Images (b) and (c) are reprinted with permission from Ref. 101, © 2016 IEEE. Image (e) is reprinted with permission from Ref. 75, © 2016 OSA. (f) Schematic illustration of the fabrication process of an $\mathrm{S}$-curved waveguide, (g) laser spectra of dual-wavelengths at 1064 and 1079 nm, and (h) RF spectrum of modelocking at 31.69 GHz. Images (f)–(h) are reprinted with permission from Ref. 159, © 2020 IEEE.

Recently, MIR ($\sim 2\mu \mathrm{m}$) waveguide lasers, typically obtained from ${\mathrm{Tm}}^{3+}$- and ${\mathrm{Ho}}^{3+}$-doped materials, have attracted much attention from researchers. For a gain medium of $\mathrm{Tm}:\mathrm{KLu}{({\mathrm{WO}}_4)}_2$, efficient waveguide lasers at $\sim 1.8\mu \mathrm{m}$ operating both in the CW and $Q$-switched regimes have been reported by Kifle et al. in various laser-written waveguide structures, such as $\mathrm{Y}$-branch splitters,²⁰⁸ surface cladding,²⁰⁵ buried circular cladding,^166,206 and photonic-lattice-like cladding prototypes.²⁰⁷ In addition, their group has demonstrated a low-loss ($<0.1\mathrm{dB}/\mathrm{cm}$) waveguide laser in a ${\mathrm{Tm}}^{3+}:{\mathrm{MgWO}}_4$ crystal at $\sim 2.02\mu \mathrm{m}$ with a slope efficiency of 64.4%.²¹¹ Also, in the multi-GHz operation regime, 5.9- and 7.8-GHz QML waveguide lasers at $\sim 2\mu \mathrm{m}$ have been successfully achieved in ${\mathrm{Ho}}^{3+}:\mathrm{YAG}$²¹² and Tm : YAG,²⁰⁹ respectively. Cr:ZnS is another host medium for MIR laser generation. Based on the cladding channel waveguide, an efficient 2333-nm waveguide laser with an output power of 101 mW has been reported by Macdonald et al.²¹⁴ Moreover, waveguides written in Fe:ZnSe crystal have shown a slope efficiency of 58%, offering a working wavelength up to 4070 nm.²¹⁵

4.2.

Frequency Converters

Frequency conversion displayed in femtosecond laser-written waveguides coming from second and higher-order nonlinearities offers an efficient manner to access special wavelengths. Similar to waveguide lasers, frequency converters in waveguide structures provide enhanced conversion efficiencies compared to bulk material, as a result of strong optical confinement and good overlap between the fundamental and nonlinear polarization waves over the whole interaction length. Femtosecond laser-written waveguides allow for nonlinear optical conversion, mainly SHG, at a wavelength range between $\sim 400$ and 790 nm.

One of the most well-studied nonlinear crystals is ${\mathrm{LiNbO}}_3$, since the work of Burghoff et al.²¹⁷ in 2007, which enabled efficient frequency doubling of 1064 nm with a conversion efficiency of 49% in laser-written dual-line waveguides. In 2021, using the regime of birefringent phase matching (BPM), waveguide geometries of multiline, vertical dual line, and depressed cladding have also been reported in ${\mathrm{LiNbO}}_3$ crystals for SHG of 1064 nm,²¹⁸ whereas BPM is difficult to realize in some crystals with weak birefringence. In recent years, LiQPM grating structures have attracted a particular interest in frequency doubling, beginning from the first proof in 2013 reported by Thomas et al.¹¹⁹ though with a poor conversion efficiency. Later in 2015, an increased conversion efficiency of 5.72% was reported by Kroesen et al., which introduced QPM grating structure inside a depressed-cladding waveguide.¹²⁷ In 2020, a waveguide-integrated 3D QPM grating scheme was proposed, as shown in Fig. 14(a), thus multiwavelength SHGs of 1065.3, 1064, 1061.6, and 1060.5 nm radiation have been demonstrated⁴³ [Fig. 14(b)]. Using this technique, Wei et al.¹²¹ produced 3D NPC in ${\mathrm{LiNbO}}_3$ crystals, as well as realizing an SHG of 829 nm with an enhanced second-harmonic power compared with the on-engineered area. Figure 14(c) illustrates the setup of the SHG process using LiQPM structures in quartz crystal, and a deep-ultraviolet generation of 177.3 nm has been achieved with a high efficiency of 1.07‰¹²⁴ [Fig. 14(d)]. In addition, based on a type I single-line waveguide, the green-up conversion at 550 and 528 nm and NIR emission at 1550 nm have been reported in $\mathrm{Er}:\mathrm{MgO}:{\mathrm{LiNbO}}_3$ crystals.¹⁴⁴ Moreover, spontaneous parametric down-conversion from 780 to 1560 nm has also been demonstrated in laser-written multiline waveguides of PPLN crystal, giving access to photon entanglement in quantum technologies.⁴⁹

Fig. 14

(a) Waveguide-integrated 3D LiQPM scheme, one period, two periods, and four periods in a ${\mathrm{LiNbO}}_3$ crystal. (b) Simultaneous SHG of four wavelengths and the fundamental and second harmonic modal profiles of a single period. Images (a) and (b) are reprinted with permission from Ref. 43, © 2020 OSA. (c) Experimental setup of the ultraviolet SHG process using LiQPM structure in a quartz crystal and (d) the SHG response signal of 177.3 nm. Images (c) and (d) are reproduced with permission from Ref. 124, CC-BY. (e) Schematic diagram of femtosecond laser-written cladding waveguide in a fan-out PPSLT crystal and (f) temperature tuning curves of seven waveguides with different poling periods. Images (e) and (f) are reprinted with permission from Ref. 148, © 2019 OSA.

Frequency doubling down to ultraviolet bands has also attracted much attention. ${\mathrm{BiB}}_3{\mathrm{O}}_6$ cladding waveguide is a good sample, and a violet SHG at 400 nm has been realized.²¹⁹ ${\mathrm{LiTaO}}_3$ crystal is another promising material owing to its shorter cut-off wavelength ($0.28\mu \mathrm{m}$) and high optical damage threshold. In 2019, a second harmonic power up to 8.5 W was realized in laser-written circular-cladding waveguides under a CW pump at 1050 nm in a periodically poled $\mathrm{MgO}:{\mathrm{LiTaO}}_3$ (PPMgSLT) crystal.²²⁰ Figure 14(e) depicts the schematic plot of femtosecond laser direct writing cladding waveguides in fan-out PPMgSLT crystals, corresponding to different poling periods. Temperature-tuned SHG for different QPM periods has been demonstrated in Fig. 14(f) with a comparable normalized conversion efficiency of $3.55\%{{\mathrm{W}}^{-1}\mathrm{cm}}^{-2}$.¹⁴⁸ Later in 2020, a tunable violet SHG over the range of 396 to 401 nm was demonstrated by varying the QPM condition in fan-out PPMgSLT crystals,²²¹ evincing efficient frequency converters based on femtosecond laser-written waveguides.

${\mathrm{KTiOPO}}_4$ (KTP) is another excellent nonlinear platform for frequency conversion. In 2007, the laser-written multiline waveguide based on type I modification permitted SHG down to 400 nm in periodically $\text{poled}:{\mathrm{KTiO}}_{\mathrm{P}}{\mathrm{O}}_4$ (PPKTP) crystals.¹⁵⁴ Lately, a further improvement of conversion efficiency ($4.6\%{\mathrm{W}}^{-1}{\mathrm{cm}}^{-2}$) has been proposed in type II dual-line waveguides in Rb-doped PPKTP crystals.²²² In particular, due to the true 3D micromachining of femtosecond laser writing, optical lattice-like $1\times 4$ beam splitters in KTP crystals have enabled SHG of 1064 nm by BPM.¹⁵⁷ In addition, the optical-lattice-like microstructures offer an efficient manner for beam mode controlling, providing a simple way to realize single-mode frequency doubling at 532 nm with a comparable conversion efficiency.¹⁸³ Table 4 summarizes the latest results for frequency converters based on femtosecond laser-written waveguides.

Table 4

Summary of latest results for frequency converters in femtosecond laser-written waveguides.

4.3.

Quantum Memories

To date, femtosecond laser direct writing has played an important role in integrated quantum information processing,^225,226 such as integrated sources of entangled photons,⁴⁹ high-dimensional transmission of quantum twisted light states,⁴⁷ 2D quantum walk,⁴⁶ and the rotated polarization directional coupler.⁴⁵ Quantum memories (QMs) are essential components in large-scale quantum networks, especially in quantum repeaters. Rare-earth ion-doped crystals are promising candidates of integrated QMs due to their large bandwidth, high multimode capacity, and high storage fidelity.²²⁷ Two approaches have been employed to manufacture integrated QMs in rare-earth ion-doped crystals, such as Ti-indiffusion in ${\mathrm{LiNbO}}_3$¹⁵ and focused-ion-beam milling.²²⁸ Nevertheless, the storage time and efficiencies in these two systems are significantly reduced. In recent years, femtosecond laser direct writing has allowed channel waveguides with improved capabilities, where the active ions are directly coupled to the light, rather than via evanescent coupling.^52,225,229 As a result, spin-wave storage with an extended lifetime, which enables long-term storage and on-demand read-out, has been achieved in waveguide-integrated QMs.^50,51,53,230 A great amount of effort has been devoted to this topic, mainly focused on the geometry of type I, type II, and ridge waveguides.

The first proof of QMs based on laser-written dual-line waveguides was reported in ${\mathrm{Pr}}^{3+}:{\mathrm{Y}}_2{\mathrm{SiO}}_5$ crystals, where the interaction with the active ions is increased by a factor of 6 compared to bulk, and demonstrated a spin-wave optical memory,²²⁹ as shown in Figs. 15(a)–15(d). Later in 2020, using the type II waveguides in ${\mathrm{Eu}}^{3+}:{\mathrm{Y}}_2{\mathrm{SiO}}_5$ crystals, on-demand light storage was demonstrated via the spin-wave atomic frequency comb (AFC) and the storage fidelity was quantitatively characterized for the first time.⁵⁰ Compared to type II waveguides, the fabrication of type I waveguides in crystalline substrates is still a very challenging task, as it requires a very narrow writing window. In 2018, Seri et al.⁵² produced high-quality type I waveguides in ${\mathrm{Pr}}^{3+}:{\mathrm{Y}}_2{\mathrm{SiO}}_5$ crystals, demonstrating efficient storage times 100 times longer than previous demonstrations [Figs. 15(e)–15(g)]. In the next year, their group reported more than 130 individual storage modes in laser-written type I waveguides in ${\mathrm{Pr}}^{3+}:{\mathrm{Y}}_2{\mathrm{SiO}}_5$ crystals, among the temporal, spectral, and spatial domains, demonstrating the capability of both time and frequency multiplexing single photon.⁵³ Such high multimodality will be necessary to realize an efficient quantum repeater scheme.

Fig. 15

(a), (b) Microscope image and near-field intensity profile of a type-II waveguide in a ${\mathrm{Pr}}^{3+}:{\mathrm{Y}}_2{\mathrm{SiO}}_5$ crystal. (c) Energy-level structure of ${{}^3\mathrm{H}}_4$ ground and ${{}^1\mathrm{D}}_2$ excited manifolds. (d) Light-storage experiments using the AFC protocol. Images (a)–(d) are reproduced with permission from Ref. 229, © 2016 American Physical Society (APS). (e), (f) End-face and top-view microscope images of type I and type II waveguides and modal profiles in ${\mathrm{Pr}}^{3+}:{\mathrm{Y}}_2{\mathrm{SiO}}_5$ crystal, respectively. (g) Time-resolved histogram for signal photons, internal storage efficiency ${\eta }_{\mathrm{AFC}}$ at different storage times, and cross-correlation values between idler photons and stored signal photons. Images (e)–(g) are reprinted with permission from Ref. 52, © 2018 Optica.

Considering that the channel waveguides (type I and type II) are fabricated at a depth beneath the crystal surface, ridge waveguide-based QMs are easily integrated with other on-chip devices, allowing for constructing large-scale quantum networks. In 2020, a laser-written ridge waveguide was successfully fabricated in an ${\mathrm{Eu}}^{3+}:{\mathrm{Y}}_2{\mathrm{SiO}}_5$ crystal, in which the properties of the ${\mathrm{Eu}}^{3+}$ ions were well-preserved.²³⁰ The spin-wave AFC storage was implemented, confirming high-interference visibility [Figs. 16(a)–16(d)]. In 2021, their group achieved a better fabrication parameter, realizing 40% end-to-end device efficiency, while the typical coupling efficiency is 10% in ${\mathrm{LiNbO}}_3$ waveguide memory. Combined with on-chip electrodes, a high storage fidelity of $99.3\%\pm 0.2\%$ and on-demand storage of time-bin qubits were demonstrated,⁵¹ far beyond that value of the classical measure-and-prepare strategy.

Fig. 16

(a) Experimental setup of coherent optical memory based on an on-chip waveguide. (b) Guided mode intensity distribution of laser-written ridge waveguides in an ${\mathrm{Eu}}^{3+}:{\mathrm{Y}}_2{\mathrm{SiO}}_5$ crystal. (c), (d) Top and front view microscope images. Images are reproduced with permission from Ref. 230, © 2020 APS.

4.4.

On-Chip Devices

Thin-film lithium-niobate-on-insulator (LNOI) has attracted significant interest in the field of PIC for its favorable low-loss limit and electro-optic coefficients.^17,18,231 With the recent revolution in fabrication techniques of on-chip photonics, great progress has been sparked in producing ultrahigh performance building blocks, such as waveguides^16,232 and microcavities,²³³ for scalable and high-density LNOI PICs. In 2015, Lin et al.²³⁴ first reported on the fabrication of high-$Q$ ${\mathrm{LiNbO}}_3$ microresonators by femtosecond laser direct writing, followed by focused ion beam milling. In 2018, their group proposed an efficient LNOI fabrication technique termed photolithography-assisted chemo-mechanical etching (PLACE).²³⁵ Figures 17(a)–17(e) schematically illustrate the process of PLACE, including mainly five steps. Notably, femtosecond laser micromachining is used to pattern the Cr thin film into a waveguide- or circular disk-shaped mask with the advantages of high scanning speed and high quality. The key in this step is to carefully choose a peak intensity of the femtosecond laser, in which the Cr film can be completely removed without damaging the underneath LNOI, since the ablation threshold of LNOI is significantly higher than that of the metal Cr.²³³ In addition, chemomechanical polishing (CMP) is employed to selectively etch smooth sidewalls capable of low scattering loss. Using this technique, a 10-cm-long LNOI waveguide with an ultralow loss of $0.027\mathrm{dB}/\mathrm{cm}$²³² [Figs. 17(f)–17(h)] and an LNOI microresonator with an ultrahigh $Q$ factor of $4.7\times {10}^7$ by improving the surface roughness of $\sim 0.115\mathrm{nm}$²³³ [Fig. 17(i)] have been demonstrated. In addition, by adding a layer of ${\mathrm{Ta}}_2{\mathrm{O}}_5$ film on the waveguides, a single-mode LNOI waveguide with a propagation loss of $0.029\mathrm{dB}/\mathrm{cm}$ at 1550 nm has also been discussed.²³⁶ Based on these ultralow-loss waveguides, Wang et al.²³⁷ demonstrated the fabrication of LNOI beam splitters and Mach–Zehnder interferometers. Later in 2020, Zhou et al.²³⁶ successfully designed and fabricated optical true delay lines on the LNOI chip, realizing a relative time delay of 2.285 ns. Various waveguides with lengths from $\sim 10$ to $\sim 100\mathrm{cm}$ have been fabricated with losses below $0.03\mathrm{dB}/\mathrm{cm}$, providing the large-scale on-chip integration of numerous unit microdevices.

Fig. 17

Schematic illustration of the PLACE fabrication process. (a) Cr thin-film deposition, (b) Cr patterning, (c) CMP, (d) chemical wet etching, and (e) coating ${\mathrm{Ta}}_2{\mathrm{O}}_5$ film. Images (a)–(e) are reproduced with permission from Ref. 236, © 2020 Chinese Physical Society (CPL). (f) Camera photo of the 11-cm-long LNOI waveguide, (g) microscope image, and (h) enlarged image. Images (f)–(h) are reproduced with permission from Ref. 232, CC-BY. (i) SEM image of LNOI microdisk. Image (i) is reproduced with permission from Ref. 233, CC-BY.

To date, some active components, such as on-chip microlasers and amplifiers, have also played an important role in the monolithic integration of PICs. In 2021, Wang et al.²³⁸ demonstrated a tunable microring laser on ${\mathrm{Er}}^{3+}$-doped LNOI, enabling a low threshold of $400\mu \mathrm{W}$. In the same year, by integrating the microelectrodes, an electro-optically tunable microring laser has been achieved on an LNOI platform on which the laser wavelength can be tuned with an electro-optic coefficient of $0.33\mathrm{pm}/\mathrm{V}$.²³⁹ Specifically, their group provided an alternative pumping manner, using an undoped LNOI waveguide directly bonded above the ${\mathrm{Er}}^{3+}$-doped microring, rather than the tapered fibers. In addition, Zhou et al.²⁴⁰ proposed an on-chip waveguide amplifier on ${\mathrm{Er}}^{3+}$-doped LNOI, which allows a net gain of 18 dB for a length of 3.6 cm emitting of 1530 nm, indicating an efficient broadband amplification.