Introduction

Flexible neural implantable systems that provide intimate and intuitive interfaces with a soft neural tissue serve as powerful tools for identifying complex neural circuits for diagnosing and treating neurological diseases1,2,3,4. Furthermore, recent efforts into simultaneously integrating multiple modalities with a single system for active recordings and actuations are of potential interest5,6,7,8. In particular, a hybrid of electrophysiological and optical approaches for neural implants maximizes the synergism of the resolution of the two methods, while at the same time compensating for the weaknesses of each approach9,10,11,12. However, in conventional device design, the two methods may conflict with each other in the process of data acquisition, such as electrophysiological readout, fluorescence cell imaging, or structural dynamics of the brain3,13,14. For example, when optical imaging, used to understand the structural and functional properties of the brain, is accompanied by opaque electrodes, there is great difficulty in analyzing the optical activation signals due to the visually obscured cells15,16,17. Even when applying optogenetics to treat neurodegenerative diseases, such as Parkinson’s disease, epilepsy, or depression, is applied to an implantable device, serious electrophysiological signal contamination due to photoelectric artifacts occurs if optical transparency is not guaranteed18,19,20,21. Transparent implantable microelectrode arrays (MEAs) can simultaneously acquire high-resolution electrophysiological signals with negligible difference in performance from conventional electrodes while alleviating fundamental problems of optical modality22,23,24,25,26,27. Unlike metal based opaque MEAs, transparent electrodes eliminate the drawbacks of multimodal systems in that they can be used to measure electrical signals from the brain in real time without optical interference through material and structural controls. Transparent, implantable electrodes are fabricated with carbon-based materials27,28, conductive polymers29,30,31, and metal nanowires32,33,34, which are primarily exploited in flexible electronics. In addition, transparent electrodes can be obtained by the geometrical design of materials for the electrodes. For instance, in the case of opaque metal, it is possible to obtain optical transparency by forming mesh-like or porous structures on the active materials, while also maintaining electrical properties like that of a conventional metal film35,36,37. Furthermore, to overcome the difficulty of securing both high electrical conductivity and optical transmittance, recent studies are focused on overcoming the disadvantages by mixing two or more materials29,38,39. Figure 1 shows the schematic overview of this paper. In this review, electrophysiological/optical diagnosis and treatment modalities are discussed, and the latest studies integrating these two approaches are introduced. Next, issues that arise when conventional opaque implantable electrode arrays are applied to optogenetics and optical imaging—photoelectric artifacts, optical image blocking, and less light transmission are the cases. Finally, a method for fabricating a transparent electrode that can solve these problems is presented by subdividing it into two categories: from a material point of view and from a structural point of view. As for the material approach of fabricating transparent neural electrodes, an introduction of intrinsically transparent material candidates, such as graphene, carbon nanotubes (CNT), conductive hydrogel, oxides, and conductive polymers, in addition to the latest research are discussed. Lastly, an introduction to research applied to biomedical engineering for a transparent electrode through a special structure is examined. This review consequently suggests the importance of transparent implantable devices for biomedical engineering and medical applications as a next-generation multimodal system for future research.

Fig. 1: A schematic introduction of the comprehensive review of the transparent neural interface.
figure 1

Schematic illustration of issues with transparent neural interface. (1) Importance of multimodal sensing & stimulation, (2) Challenges from non-transparent electrode, (3) Design method for transparent neural interface.

Full size image

Electrical & optical modalities

Recently, implantable neural devices that can identify complex neuronal circuits and diagnose and treat neurodegenerative diseases have been extensively studied40,41,42. Among them, the electrophysiological approach is a representative method for analyzing neuro-dynamics and neurological disorders43,44. The biggest advantage of an implantable device for electrophysiological recording is that it exhibits high temporal resolution as it can be exploited to measure biological signals in real time45. This property is advantageous for pathological diagnosis due to its potential for sophisticated analysis of signal transmission, even from a single neuron46. However, the methods of inserting devices into the brain or other organs are spatially limited because they can be operated only at the implant site. From an electrophysiological point of view, ‘spatial resolution’ varies greatly depending upon the size and spacing of electrodes for measuring neural activity. Especially, size of the electrophysiological sensing site is a significantly important factor in the performance and sensitivity of the neural interface. For example, although it has an advantage of being able to measure the activity of individual neurons in proximity with a small recording site (<1000 um2), it is unavoidable from degradation of signal quality due to an increased electrochemical impedance. Therefore, for high spatial resolution electrophysiological recording, the size of the electrode densely distributed on a tissue should be reduced as much as possible to increase selectivity maintaining a low impedance value41. In other words, a device without empty measurement sites with dense electrodes in the same implanted area has high spatial resolution. In the case of a passive electrode array composed of one channel per one conductive interconnect, dense arrangement of recording channels is difficult due to wiring spatial charge for signal transmission. This is the reason that the spatial resolution of the electrophysiological approach is lower than that of the optical approaches47. To overcome this weakness, as shown in Fig. 2a, devices comprised of high-density, multi-channel electrode arrays are being thoroughly studied48,49. However, it is not entirely free from spatial constraints compared to optical approaches, such as magnetic resonance imaging (MRI), computed tomography (CT), or positron emission tomography (PET), which structurally analyze the brain or target organs.

Fig. 2: Integration with electrophysiology and optical modalities by neural implantable device.
figure 2

a High-density active electrode array for electrophysiological brain mapping (left) and a high-density, active-matrix neural probe for deep brain recording (right). b Transcranial Ca2+ transient imaging from a visual flash response of anesthetized mouse (left) and total hemoglobin imaging of a mouse cortex after hind-paw stimulation (right). c Graphene fiber-based neural electrode for electrical stimulation of subthalamic nucleus (left) and endovascular stent for focal electrical stimulation of the sheep motor cortex (right). d Soft, fully implantable optoelectronic system for light stimulation of nucleus accumbens (left) and ultrasoft-hydrogel optical fibers for optogenetics stimulation (right). e Stretchable electrocorticography (ECoG) array for simultaneous electrophysiological recording with optogenetics stimulation. f Two-photon fluorescence imaging of GcaMP6f-labeled neurons from a SNAP25-GCaMP6f mouse (left) and corresponding calcium transient from each cell in colored circles from imaging and ECoG signals from the electrode (right). g Three-dimensional multi-shank probe for simultaneous electrical modality, optical modality, and drug delivery. A magnified image of the device is presented with an optical waveguide (left). The connectivity from a first somatic region to a second somatic region is illustrated with a three-dimensional structure of the multi-shank probe (right). Panel a reprinted with permissions from ref. 48,49, AAAS (left) and Springer Nature (right). Panel b reprinted with permissions from ref. 50,51, Springer Nature (left) and Elsevier (right). Panel c reprinted with permissions from ref. 77,78, Springer Nature. Panel d reprinted with permissions from ref. 81,82, Springer Nature (left) and Wiley (right). Panel e and f reprinted with permissions from ref. 85, Wiley. Panel g reprinted with permissions from ref. 86, Springer Nature.

Full size image

In addition to the brain structure imaging approaches mentioned above, several optical approaches can observe brain activation on a cell-by-cell basis50,51 (Fig. 2b). Optical imaging approaches are largely divided into methods of molecular selective neuro-dynamics identification using fluorescent52,53,54,55 and voltage-sensitive dyes, and methods analyzing intrinsic scattering and oxygenated condensation changes in brain tissue such as optical coherence tomography (OCT)56,57 and diffuse optical tomography (DOT)58,59. An advantage of using optical recording is that it shows a better spatial resolution compared to the electrophysiological approach because the desired site can be observed broadly without a spatial disruption of the implanted site. Voltage sensitive dye imaging is a method of measuring neural activity using molecules that can modify electric charges from neural circuits into fluorescence of emitted light. Since this method provides only the surface activity of the brain in nearly two dimensions, it is difficult to see the activity of the deep brain relative to the electrophysiology through the penetrating probe60,61,62. However, this approach is particularly important when neural activities occur in many cells or multiple sites simultaneously, such as in the nerves or brain. It is also advantageous when observing cells that are too narrow for electrophysiological recording or cells prone to damage. Particularly, calcium transient fluorescence of neurons or HbT imaging, which is a method measuring changes in oxygen saturation of hemoglobin, are representative methods to study the spatial activity of neuronal populations63,64,65. These approaches show excellent temporal resolution of up to 3–10 ms depending on the tissue staining method or the resolution of microscopy, but they require integration with near-real-time electrophysiological recording methods (~1 ms) to function as clinical applications17,66.

The method of ‘treatment’ beyond the simple ‘diagnosis’ of neural activity is also important in terms of the functionalities of implantable devices. There are several ways to modulate brain nerve activity, including electrical stimulation67,68, light stimulation69,70,71, drug delivery72,73,74, and magnetic stimulation75,76. Among the electrophysiological treatments, the most representative approach is electrical stimulation, which is the process of modulating the neuronal activity and firing patterns through current or voltage by inserting an electrically conducting microelectrode into the body77,78 (Fig. 2c). Electrical stimulation has been already widely used clinically for decades as its therapeutic effect is guaranteed. However, this is limited as local stimulation to a specific area is impossible due to the undesired excitation of neighboring neurons79,80. To overcome this issue, optogenetic stimulation that can precisely control the neural activity of target neurons through light has been rigorously studied81,82 (Fig. 2d). Optical stimulation exhibits excellent selectivity to neurons because it changes neuro-dynamics by attaching light-sensitive proteins, such as channel rhodopsin (ChR2) or halo rhodopsin (HR), and then applies a specific wavelength of light. This method allows versatile, cell type-specific light-based control of cellular activity in diverse set of cells, circuits and brain structures than that of an electrical stimulation.

A multimodal implantable device is a neuro-prosthetic that can be applied as a single tool that integrates various methods of both recording the above-mentioned neural activity and treating diseases. Neural electrode array for electrophysiology is largely divided depending on the site of implantation in the brain-electrocorticography (ECoG) and penetrating probe. In particular, ECoG has an advantage of capturing neural activities of epilepsy or neuro-degenerative disorder more broadly than that of penetrating probe. Since the ECoG is a method of measuring the integrated activity (field potentials) of thousands to millions of nerve cells, the signals from the ECoG has become an important clue pathological diagnosis83,84. The hybrid system of the electrophysiological and optical approaches maximizes the advantages of each method and complements the disadvantages of each method. Figure 2e shows the device configuration of a study in which two-photon calcium imaging and electrocorticography were simultaneously performed85. Fluorescence imaging of a large field of view helped analyze the neuronal activity more intuitively with an electrical signal recorded from the limited measurement of the electrodes (Fig. 2f). However, shadows from the recording electrode in the optical image imply that it was not completely free from optical blocking by the non-transparent electrode array. Additionally, the lack of a temporal sampling rate and the inaccuracy of fluorescence changes, which are main problems of the imaging methods, can be solved by ECoG recordings. Similarly, several attempts have been made to design a device that can bridge the gap between optogenetic and electrophysiological recordings. For example, there is a multi-shank probe that can be used to diagnose neural propagations transmitted through light stimulation by electrophysiological action potential analysis86 (Fig. 2g). This 3D neural interface confirmed the functional connectivity of the two somatic regions through synaptic latency and transmission velocity while simultaneously analyzing the neural network by optical stimulation.

Challenges from opaque electrodes

The most important part of combining electrophysiology and optical modality is that the data from the two methodologies should not be polluted the other and should generate diagnostic or therapeutic synergies with their own device characteristics. Recent studies have revealed that opaque electrodes based on conventional metals, such as gold or platinum, present new challenges and boundaries in the integration of electrophysiological and optical modalities. In this section, we discuss various problems that arise from multimodal diagnosis and treatment using opaque neural interfaces: photoelectric artifacts, optical image blocking, and light transmittance efficiency. Additionally, we also present how transparent devices can alleviate these problems. A summary of each issue is shown in Fig. 3a.

Fig. 3: Challenges with the integration of electrophysiological and optical modalities by non-transparent neural interfaces.
figure 3

a Schematic illustration of challenges from non-transparent electrode array with integration of optical modalities. b Orbital electron energy levels versus the density of each state. c Photoelectric artifact from electrophysiological signals from the brain after 473 nm photo-stimulation with an optical fiber. d The trace of electrophysiological signals from panel c. The photoelectric artifact is presented as the onset of stimulation. e Optical Coherence Tomography (OCT) imaging of a brain with a non-transparent platinum electrode array (left) and a transparent graphene electrode array (right). f An image of blood vessels from the brain cortex labeled with rhodamine-B dextran through the platinum electrode array from the cranial window. g Fluorescence image of blood vessels with a transparent graphene electrode array. h Spatiotemporal electrophysiological heatmap of a corresponding electrode channel upon illumination of the 473 nm laser directly on the surface of the Au film electrode (left), and an Au nanonetwork electrode (right). i Spatially distributed electrophysiological potentials from the optogenetic stimulation. The blue color represents the illumination spot. Panels b, c, and d reprinted with permissions from ref. 19, Royal Society of Chemistry. Panels e and g reprinted with permissions from ref. 16, Springer Nature. Panels f and i reprinted with permissions from ref. 90, IOP publishing. Panel h reprinted with permissions from ref. 18, Wiley.

Full size image

Photoelectric artifacts

In the field of neuroscience, the photoelectric effect generally refers to a phenomenon in which a potential difference occurs in an electrode when a metal electrode receives light and current flows at the surface of the electrode18,19. When diagnosing neuro-dynamics optically, such as fluorescence calcium imaging, the surrounding residual current does not have a significant effect on the cell. However, when the electrophysiological signal by optogenetics is recorded, the current around the electrode surface caused by light energy is the main source of data contamination and distortion. A photoelectric artifact in a metal-based microelectrode occurs only if it exceeds a specific minimum photon energy, termed the threshold wavelength, which can emit photoelectrons. Thus, threshold frequencies or photons have enough energy to emit electrons, depending on the proper material choice for the device. Figure 3b is a schematic diagram of the orbital energy level of an electron. This schematic suggests that the orbitals of all solid materials have bands of specific energy ranges. For example, the conduction band refers to the state in which electrons can escape from their atoms or molecules in the form of a free charge, and the valence band describes the region in which electrons are in the lowest energy state. In the case of metals, since the valence and conduction bands overlap with one another, there is almost no bandgap, so electrons are relatively easy to change into free charges. Since the semiconductor material has a small band gap, electrons can be delocalized into a conduction band by applying a certain amount of heat or light energy. In the case of an insulator, electrons cannot easily reach the conduction band by a small amount of heat or light energy, as it has a very large bandgap. The metal-based microelectrode has a high probability of causing photoelectric artifacts because the photo excitation of metal electrons is significantly easier than those of other materials87,88. Figure 3c shows electrophysiological signals contaminated by photoelectric artifacts. Photoelectric artifacts caused by laser pulse stimulation show similar tendencies to single-unit action potential waveforms. In the case where the amplitude of the voltage significantly increases at the start of light irradiation, it is easy to recognize that it is an onset noise caused by the light source. It is possible to minimize this problem through signal wave filtering or linear regression of the trace of the signal, as presented in Fig. 3d. However, for the signal that follows, it is unclear whether it is noise caused by photoelectric artifacts or a biological signal from an experimental model, so it has a great influence on the interpretation of electrophysiological results. To overcome this issue, optical stimulation of the brain was dominated by low-frequency or long rise time waves to reduce light-induced artifacts. These limitations frustrate the possibility of treating neurological diseases that require high-frequency light stimulation. In conclusion, to enable light stimulation in a wide frequency band without interference by artifacts, it is necessary to select an electrode material that hardly generates a photoelectric effect or to design a device with a special structure89.

Optical image blocking

Another issue that occurs when opaque electrodes are used for neural interfaces is tissue imaging occlusion by the electrodes. Brain imaging is a method of observing activated tissue with a high spatial resolution, without damaging cells by using a specific band of light. Beyond structural imaging of the brain, the biggest drawback of this functional imaging is that obtaining neurodynamic information with a high temporal resolution is difficult. To overcome this challenge, an integration of electrophysiological approaches that ensure real-time biological information is significantly important. Figure 3e shows an OCT image of a metal-based opaque electrode (left) and an OCT image of a transparent electrode (right)16. To analyze the electrophysiological signals from the brain, the implantable device is contacted on the surface of the tissue, and the opaque electrode directly covers the blood vessel and brain tissue to be observed. This problem not only impedes the interpretation of signal propagation in functionalized tissues, such as with voltage-sensitive dyes, but also makes it difficult to observe vascular changes and the dynamic activity of nerves. The weaknesses of using an opaque neural interface are deteriorated as the electrode channel density increases. In a passive device consisting of only conductive connections, the number of channels for recording electrodes and the size of the electrodes are severely limited because the number of connections required per electrode increases proportionally, as shown in Fig. 3f90. Conversely, in the case of a device in which all connections, including electrodes, are transparent, cell-specific activity can be confirmed without optical hindrance (Fig. 3g)28.

Less light transmission

Brain stimulation through optical modality is affected by frequency and light intensity91,92,93. Therefore, it is evident that the neural interface, which is an intermediate pathway from the light to brain tissue, should not interfere with the light to reach the nervous tissue without significant energy loss. Figure 3h shows the light transmission efficiency of a transparent electrode by representing the optical stimulation from electrophysiological distribution using a pristine gold film electrode and a gold nanonetwork18. Optogenetic stimulation was performed directly above the surface of the electrodes, and, as the light intensity is the same, this suggests that the transparency of the electrode array critically affects light transmission. Schematics of brain color-mapping of light stimulation on the electrocorticography (ECoG) microelectrode array from gold film shows the electrode array absorbs most of the light, clearly demonstrating the failure of stimulation on a transgenic mice. Conversely, a gold nano-network electrode array showed the propagation of a relatively high amplitude potential from the same stimulation site to the surrounding area, proposing the possibility of significantly increasing the light transmission efficiency with the same material. Alternatively, as shown in Fig. 3i, the stimulation site can be set between the electrode arrays upon light illumination90. However, the expected electrophysiological signal cannot be obtained from the target neuron of the light stimulus when the position of the neuron is located just underneath the opaque electrode as it completely blocks the stimulus. The most ideal approach to overcome the aforementioned problems is to design electrodes through the engineering of materials and structures, thereby establishing user transparency of the materials. In the next section, we discuss the fabrication method of transparent electrodes by material and structure and its latest research.

Transparent implants-from materials

This section discusses the fabrication of electrodes for neural interfaces using intrinsically transparent materials. Two representative methods of transparent neural interface fabrication utilize (1) intrinsic transparent materials and (2) structural modification of opaque materials. A detailed introduction of transparent material candidates is presented in Table 1. Transparent neural interfaces must ensure transparency to minimize light-induced artifacts, as well as conductivity to measure electrical signals3,94. To transmit neural signals properly, excellent electrochemical impedance between the electrode sensing site and tissue is also important. Moreover, the trace resistance between the sensing site and the percutaneous connecter becomes a dominant factor in obtaining high-quality electrophysiological signals. Assuming the same electrode interconnect line width and length, carbon or polymer-based electrodes have much more unfavorable electrical properties than metal wiring in terms of trace resistance. Therefore, in the case of a carbon-based neural electrode array, the interconnect line is mainly substituted with a conductive metal to make the electron path more intuitive, while giving up interconnect transparency. For polymer electrode arrays, the trace resistance is significantly reduced by changing the molecular structure through special doping or post treatment, maintaining the transparency of the entire device including conducting path41,95. Therefore, high transparency and conductivity should be the priority consideration when selecting materials for the design of multimodal devices. Recently, graphene has attracted much attention due to its advantageous properties associated with neural interfaces. Beyond its superior electrical conductivity, it also has excellent transparency due to its two-dimensional honeycomb structure. Moreover, its well-known mechanical characteristics, such as high tensile strength and high flexibility, also make graphene a promising material for reliable and robust flexible electronics. In this section, detailed properties and related studies on graphene are presented. Moreover, carbon nanotubes, which are rolled-up graphene sheets, conductive hydrogels with tissue-like Young’s modulus, and transparent conducting oxides with low resistivity are also discussed.

Table 1 Summary of material candidates for transparent neural interface.
Full size table

Transparent graphene neural interface

Graphene, one of the most promising candidates for transparent neural interface fabrication, is a carbon-based 2D material arranged with a honeycomb-lattice structure3,96,97. Carbon-based materials are basically black soots which represents low light transmittance in bulk state. However, owing to negligible thickness of 2D monolayer graphene, which is much smaller than wavelength of the visible light, ideal 2D monolayer graphene shows tremendous visible light transmittance over 97%.98,99. Moreover, graphene has a superior intrinsic optical transparency of over 90% that covers ultraviolet (UV) and infrared (IR) light16,100. High visible light transmittance, which greatly reduces optical blocking, is critically important in transparent neural interfaces. Furthermore, high optical transmittance of the neural interface materials for both UV and IR light is highly advantageous in terms of optogenetic stimulation and photo-induced imaging98. Above all its great optical properties, this 2D monolayer material performs outstanding mechanical and chemical stability because of strong sp2 bonding between carbon atoms96,97,101. For example, it has durable physical properties, such as a breaking strength of 42 N m−1 and a Young’s modulus of 1.0 TPa102. In addition, as it features high electrical mobility, graphene is excellent material for flexible or stretchable electronics103,104,105,106,107. In this respect, recently, graphene-based flexible electrodes have been widely utilized as bio-integrated transparent electronics.

Most recent studies using intrinsically transparent materials utilize graphene as transparent electrodes owing to its outstanding electrical, mechanical, and optical properties10,15,24,28,108,109. Although graphene is not completely free from photoelectric artifacts, it shows powerful performance for an optical approach than metal electrodes with the low probability of photon energy excitation28,110. By using its broad wavelength transparency, a similarity between calcium imaging and electrophysiological signals was verified. Figure 4a shows a schematic illustration of a transparent neural interface using graphene. The transparency of the electrode is owing to the honeycomb-lattice structure of graphene as we can see in the figure24. Figure 4b illustrates fluorescent calcium images under the graphene electrode on a P12 mouse hippocampal slice using a 488 nm excitation wavelength. Modification of ion concentrations, such as potassium and magnesium ions, and bicuculline-methiodide were utilized to make spiking activities of the sample. Simultaneously, electrophysiological signals were recorded by the electrodes without noticeable laser-induced artifacts, and it is shown together, as illustrated in Fig. 4c, with calcium transients of the numbered cell. This implies that a strong correlation exists between fluorescence transients and electrophysiological signals using transparent graphene electrodes.

Fig. 4: Graphene-based transparent electrode array for simultaneous electrical and optical modalities.
figure 4

a Schematic of a graphene-based transparent neural interface. The honeycomb-lattice structure of graphene maximizes electrode transparency. b Calcium transient images under a graphene-based transparent electrode with 488 nm excitation wavelength. Six cells are numbered in the region of interest. Scale bar: 50 µm. c Correlation between electrophysiological signals reflecting manually induced interictal-like effects (top) and simultaneous ΔF/F0 curves for numbered cells in panel b (bottom). All data from panels b and c are observed simultaneously. d Fluorescence image of a GCaMP6f mice cortex under a transparent graphene electrocorticography (ECoG) array. Scale bar: 200 µm. e Spatiotemporal fluorescence recordings with various electrical stimulation amplitudes. f Plot of intensity changes versus the lateral position. The greatest intensity changes were detected at the stimuli point (between 1.5 and 2 mm). The lateral position axis corresponds to the horizontal red line in panel e. g Bright optical image of the transparent graphene electrode array on the brain surface. Scale bar: 500 µm. h Two-photon imaging at a depth of 1,200 µm under photoelectric artifact minimized graphene electrode. Scale bar: 100 µm. i Crosstalk-free detection of two-photon imaging and electrophysiological recording, according to simultaneous optogenetic stimulation. Panel a, b, and c adapted from ref. 24, Springer Nature Limited. Panel d, e, and f adapted with permission from ref. 15, American Chemical Society. Panel g, h, and i adapted from ref. 109, Springer Nature Limited.

Full size image

In general, graphene-based transparent neural interfaces requires a high quality graphene but the transfer process of the graphene on a subtrate make the fabrication complexity. In addition, light absorption and dissipation by the substrate site are critical in optical modalities. To maximize transparency of the electrode, a Parylene C and graphene-based transparent electrode was fabricated. A fluorescence image using the device is shown in Fig. 4d15. Charge density limitation for the four-layer graphene electrode was characterized based on EIS, CV, and a failure bench-testing method. Spatiotemporal responses with varying stimulation currents in the calculated range (approximately 150 μA) verified the correlation between the stimulation current and neural activation represented as a fluorescence transient, as shown in Fig. 4e. Also, spatial response analysis proved that maximum activations of cells occur at the center of the stimuli (between 1.5 mm to 2 mm) as shown in Fig. 4f.

The capability of minimizing light-induced artifacts is a key feature of transparent neural interface. Although high-quality graphene produces negligible artifacts, ordinary graphene electrodes bring out unexpected artifacts due to defects or residues from the fabrication process. In this regard, graphene transparent electrodes with highly minimized artifacts were formed by improving the fabrication process. Figure 4g shows a bright field image of a 16-channel electrode device without any visible electrode sites109. In addition to the visible light transmittance, highly minimized artifacts allow deep two-photon imaging below the cortical surface under transparent electrodes10. Nevertheless, a two-photon excitation laser induces unnecessary artifacts which impede clear imaging, and the 1200 µm deep imaging of tissue beneath the electrode was successfully obtained as shown in Fig. 4h. Also, two-photon imaging, optogenetic stimulation, and large-scale hemodynamic response imaging were combined with electrophysiological recordings of the cortical surface without any noticeable interruptions. By using these crosstalk-free electrodes, simultaneous electrophysiological recording, and two-photon imaging of the cortical surface during optogenetic stimulation was achieved, as shown in Fig. 4i.

Other promising transparent implants

Transparent neural interfaces using intrinsically transparent materials other than graphene are also being widely investigated. Carbon nanotubes (CNTs), another highly conductive carbon-based material, is one of the more powerful candidates. A single CNT is rolled-up graphene sheet into a tube shape and is categorized as single-walled CNT (SWCNT) and multi-walled CNT (MWCNT), based on the number of graphene sheets104,111,112. A single CNT could have metallic or semiconductor properties depending on its chiral vector. In general, CNTs are a mixture of both metallic and semi-conducting CNTs, and their ideal separation is almost impossible due to van der Waals force attractions113,114. Therefore, the properties of CNTs are determined by adjusting their ratios. CNT electrodes maintain their conductivity through nanowire percolations. These nano mesh-like void networks and the electrode thickness far under visible light wavelength enable high light transmittance of CNT electrodes99,115. Even though CNTs are widely used to make transparent electrodes, optical incident orientation should be examined because of CNTs optical anisotropy. It is because of 1D structure and nanometer-scale diameter of CNT, which makes different optical strengths between tangential and parallel direction transitions based on its axes. In general, transmittance of CNTs electrode decreases as incident angle increases, so such optical property should be carefully considered116,117. Interestingly, in the CNT networks, the conductivity is limited by the outmost shell, even if the number of graphene sheets increases. Therefore, as the number of shells increases, a greater amount of light is absorbed, resulting in less light transmittance when the same conductivity is desired. As such, SWCNTs-based electrodes perform greater optical properties compared to MWCNTs electrodes in the same fabrication conditions99. In addition, its massive electrical conductivity and mobility make CNTs a highly applicable material for transparent neural interfaces97,107,118. Since CNTs can be dispersed into a variety of aqueous solutions, roll-to-roll fabrication, which has great a merit in industrial manufacturing, is possible119.

In addition, the possibility of easy fabrication processes, such as solution-based spray coating, is a strong advantage compared to a method of forming transfer-based graphene electrodes120,121. A recent study shows highly stretchable transparent neural interfaces by patterning CNT film on an elastomer. Figure 5a displays a schematic illustration of the device and its functionality27. Carbon nanotube percolations on stretchable and transparent substrates made the device stretchable up to 20% with an electrical impedance increment of 26%, maintaining its high optical transparency. Also, negligible light-induced artifacts were confirmed compared with those of opaque gold electrodes, as plotted in Fig. 5b. Conventional flexible neural interfaces have difficulty to maintain intimate contact with the complex curvilinear shape of the brain surface. However, stretchable properties of the hydrogel electrode enables successful measurement of electrophysiological signals on complicated brain surface during seizure-like ictal discharge. At the same time, as the artifacts were minimized, the association between the ECoG signal and two-photon calcium imaging during ictal discharge was successfully confirmed, as shown in Fig. 5c.

Fig. 5: Additional transparent materials for simultaneous electrical and optical modalities.
figure 5

a Carbon nanotube (CNT) transparent electrode array on a stretchable substrate. b Minimized light-induced artifact of a CNT electrode compared with a gold electrode. c Simultaneous electrode recording (top) and calcium transients under the CNT electrode (bottom) during seizure-like events. Two-photon images for dashed timelines on calcium transients (right). Scale bar: 10 µm (right). d Schematic illustration with highlighted features of a hydrogel-based transparent electrode. e Transmittance versus wavelength plot for hydrogel-based electrode materials. f Hydrogel-based and conventional electrode on the complicated surface of porcine brain. The hydrogel electrode performs better in terms of transparency and conformality to the surface. Scale bar: 10 mm (left), 5 mm (right). g 4 × 4 Zinc-Oxide-based (ZnO) transparent micro-opto-electrode array (MOA). Optogenetic stimuli was transmitted through the ZnO electrodes and emitted from the tip as waveguides. These pinpointed stimulations minimize spatial mismatches between the electrical and optical modalities. Scale bar: 1 mm. h High-powered optical stimulation to the motor cortex prompted various body movements. In the case of optoelectrode 9, both forelimbs reacted and corresponding electrical recordings are plotted. i Electrical recordings from an MOA when various powered optical stimulations were given to optoelectrode 16. j Transparent PEDOT:PSS-EG neural electrode array for simultaneous electrophysiology with optogenetics. k 473 nm light stimulation upon the electrode of Au array (Top) and PEDOT:PSS-EG array (Bottom), respectively. l Electrophysiological color-heatmap from light evoked signal response with conventional Au electrode array (Top) and transparent PEDOT:PSS-EG electrode array (Bottom). Panel a, b, and c adapted with permission from ref. 27, American Chemical Society. Panel d, e, and f adapted with permission from ref. 129, Springer Nature. Panel g, h, and i reprinted from ref. 144, Springer Nature. Panel j, k, and l adapted with permission from ref. 154, Springer Nature.

Full size image

Biomedical devices using hydrogels, hydrophilic polymers which have similar softness to living tissues (tens kPa), are highly conformable to complicated surfaces due to their soft nature. These properties are not only beneficial for maintaining conformal contacts, but also effective to preventing tissue damages due to Young’s modulus inequality104,122,123. Hydrogels can have a variety of electrical, optical, and mechanical characteristics, depending on the materials being combined. Many notable studies have been conducted such as implementing anisotropic conductivity or wet surface adhesion using hydrogels122,124,125,126,127. Among them, conductive hydrogels with high transparency have been considered as an active material for transparent neural interface fabrication128,129. For example, a polyvinyl alcohol (PVA) hydrogel can achieve high transparency when PVA powder is mixed with dimethyl sulfoxide (DMSO) in specific proportions. In the synthesis of transparent PVA hydrogels, hydrogen bonding between water molecules and DMSO and the crystallinity of PVA hydrogel are strongly related130. Going deeper, the larger the crystalline volume, the less transparent the hydrogel as it tends to block the penetration of light. The hydrogen bonding of PVA plays the most important role in this crystalline form. If the PVA hydrogel is synthesized in pure water only, hydrogen bonds are formed between the PVA molecular chains. The remaining water molecules are then crosslinked by bonding with the -OH group of the PVA molecular chain, promoting crystalline structure growth in the z-dimension, increasing the crystalline volume. Therefore, reducing the crystalline volume is a key to form a highly transparent PVA hydrogel, and a 1:2 DMSO to water network structure fulfills this role. When the concentration of DMSO is more than 40% by weight, the oxygen atoms of DMSO bond with hydrogen molecules of water molecules to form a 1 DMSO:2 H2O formation, reducing the hydrogen bonding between the PVA molecular chain and water molecules. Meanwhile, PVA crystals grow two-dimensionally, with a lower volume of the PVA crystalline structure yielding high transparency. An ion-interfacing transparent neural interface, which utilizes the conformity and transparency of the PVA hydrogel, has been developed. Figure 5d shows the device schematics and highlights features of the device129. Unlike many other neural interfaces, which are generally implemented in plastic films, the hydrogel device minimizes disturbances of tissue fluid circulation by using a PVA hydrogel and PDMS, which have high water permeability compared to those of plastic films (e.g., polyimide)131,132. In addition, the electrode site and the recording site were separated using a salt bridge (KCl-saturated PVA) to implement high transparency, and light transmission, according to the wavelength band, is plotted in Fig. 5e. This transparent ion-interfacing hydrogel device successfully measured the ECoG signal of porcine brain without optical blocking. Figure 5f shows superior conformality and transparency of the device compared with conventional electrodes on the complicated cerebral cortex surface of the porcine brain.

Transparent conductive oxides (TCOs) are doped oxide compounds including metallic elements with both sufficient electrical conductivity and optical transparency133. The conductivity of TCOs is due to extrinsic dopants and oxygen vacancies. Without suitable doping, these oxides usually work as insulators. In general, optical band gaps of TCOs are significantly wide (>3 eV)134,135. TCOs show a remarkably small amount of light absorption between near-UV and near-infrared range (NIR) owing to this wide optical band gap. However, rapid transmittance level drops occur in the UV range due to strong energies of the UV light exceeding the optical band gap of TCOs. Also, optical characteristics of the NIR region is degraded mainly because of plasma edge reflections. In general, the optical transmission windows are interconnected with the conductivity of TCOs. This is because correlations exist in TCOs between carrier density and light absorption rate based on the Moss-Burstein effect133,135. Indium tin oxide (ITO) is the most widely used TCO for transparent thin-film fabrication. It exhibits high visible light transmittance of over 80% and sufficient electrical conductance. Although ITO has great optical and electrical properties, ITO is less likely to be considered as an active material for transparent neural interfaces due to its brittleness and poor manufacturing cost effectiveness97,136,137. Therefore, recent efforts to develop alternative transparent oxide-based materials to replace ITO have been made136,137,138,139,140,141,142,143. Due to the absence of high-performance electro-optical multimodal probes, multi-site optical stimulation with simultaneous electrical recording for microscale spatiotemporal resolution has not been well developed. In ordinary transparent neural interfaces, inevitable spatial mismatches occur between electrical recording and optogenetic stimulation points. These unavoidable differences were successfully minimized with negligible photoelectric artifacts using an ITO-coated ZnO transparent neural interface. Figure 5g shows a 4 × 4 micro-opto-electrode array (MOA) which transmits optical power to exact points of the electrical recording sites in waveguide format144. Owing to the capability of the pinpointed stimulation, a functional map of the motor cortex was verified. For example, Fig. 5h shows electrically measured firing rates from intense light stimulation of opto-electrode 9, which is colored blue in the figure. Simultaneously, both forelimbs reacted to the stimulation. In addition, it was confirmed that the size of the tissue responding to the optical stimulation varies with stimulation intensity. Figure 5i shows the stimulated region broadens as the intensity of light increases when opto-electrode 16 is manipulated.

Conductive polymers are used in various ways in implantable devices from their high mechanical flexibility that leads to reducing the risk of mechanical mismatch between the device and the soft tissues thanks to its low Young’s modulus, and versatile change of electrical properties due to diversification of synthetic methods145,146. The excellent electrical properties of conductive polymers are attributed to the π-π conjugation molecular structure in which single and double bonds of molecules alternate147. Electrons belonging to this π-conjugated structures are delocalized in atoms along the π-π bond, consequently, make them free electrons. Conductive polymers have dramatically different electrical properties depending on the oxidation and reduction states of each material148. This is because the π-π bond exists in an oxidized state to represent a positive charge, and the dopant for neutralization of the charge coexists with the polymer in the form of an anion. Promising conductive polymers exhibiting excellent electrical conductivity have been broadly applied to implantable devices including polypyrrole (PPy)149, polyaniline (PANI)150, and PEDOT:PSS (poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PSS)). Among them, PEDOT:PSS represents intrinsically high optical transmittance and has the potential to have high electrical conductivity depending upon the post treatment151. These characteristics have been mainly used as an adhesion layer to reduce mechanical mismatch of metal-based implantable microelectrodes and to secure low electrochemical impedance152,153. For an example, Fig. 5j shows a schematic illustration of a transparent implantable electrode array composed of a single layer of PEDOT:PSS treated with ethylene glycol154. The device is highly transparent including the substrate, and this particular device for neural recording and neuromodulation by combining with optogenetics is manufactured by the simplest fabrication. Au electrode array and transparent PEDOT:PSS-EG electrode array were implanted over the right cerebral cortex of transgenic mice to figure out how much blue laser was transmitted through the electrode to the brain (Fig. 5k). As shown in Fig. 5l, the conventional Au electrode array showed significantly lower light evoked electrophysiological potentials due to low optical transmittance, whereas the PEDOT:PSS-EG electrode array accurately transmits light to the laser stimulation point. Traditionally, electro-plated PEDOT:PSS on metal electrodes or other conductive electrodes are used to reduce the impedance for neural recording152. However, the method of electroplating PEDOT:PSS has some complexity in the manufacturing process because the deposition of an additional conductive layer for the adhesion layer is essential. In addition, due to the intrinsic transparency of this metal layer, the deposition thickness of PEDOT:PSS is also limited to ensure the transparency of the entire electrode array. Herein, by dramatically improving the post treatment method on single layer PEDOT:PSS for enhancing its conductivity and simple lift-off process, the manufacturing cost for the transparent electrodes is significantly reduced.

Transparent implants-from structure

Although there are methods that use an intrinsically transparent material, as introduced above, to design the transparent MEAs, there is another recent research trend of modifying existing materials to have optically see-through properties through structural design. One of the significant advantages of this approach is that it maintains optical transparency while minimizing the loss of the electrical properties of each material. In the following section, recent studies into various mesh structures and metallic nano wires will be introduced as two promising candidates of the structural design.

Mesh structured neural interface

Mesh structure is one of the alternatives to design transparent metal interfaces with intrinsically non-transparent materials. The optimum performance of the mesh-structured electrode lies in balancing sheet resistance (RS) and transparency155. Two factors of this structure can be modified by adjusting various mesh values (width, pitch, thickness) or the geometry of the patterns. The sheet resistance varies depending on the design of the geometrical pattern (grids, holes, pyramids, honeycombs, etc.), taking a square grid as an example, the expression for calculating Rs can be expressed as below where ρ represents specific resistance while others, h (height), w (width), and p (pitch), are mesh values. Transparency (T) can be predicted from these mesh values under the premise that the geometrical design is a grid. By substituting the mesh value with the appropriate value, the relationship between sheet resistance and transparency can be expressed by the following equation35,156:

$$R_{s,grid} = \frac{{\rho \cdot p}}{{h \cdot w}} = \frac{\rho }{h}\left( {\frac{{1 - \sqrt T + T}}{{1 - \sqrt T }}} \right)$$

As shown in the expression, the wider the spacing between the metal wires, the higher the transparency, along with the unavoidable high electrochemical impedance. Several studies of designing a transparent MEA using a mesh structure have used a method of coating or bonding low-impedance materials, such as conductive polymers or oxides, on a metal grid to avoid the disadvantages157,158,159,160,161. Figure 6a shows a highly transparent conducting microelectrode array based on a metal grid (MG) hybrid with ITO162. Due to the low impedance and high transparency of ITO, placing the ITO in the vacant space of the MG significantly increases the charge transfer and interfacial area between the electrode and the target region without loss of transparency. The combination of ITO and the MG is possible both ways, either by adding ITO to the MG in an island form or by laying an ITO film under the MG. In the case of a hybrid microelectrode based on the ITO island type, the MG structure bridges the gaps in ITO to achieve highly flexible hybrid microelectrodes, structurally minimizing the mechanical load applied to the intrinsically brittle ITO islands. This neural microelectrode showed high transparency at 81 ± 1.9% from 400 to 800 nm with sheet resistance cut in half with 14.1 Ω sq−1 compared to pristine MG microelectrodes (26.8 Ω sq−1). A demonstration of the electrophysiological signal was measured of the hearts of transgenic mice in vitro when stimulated with blue light LEDs (Fig. 6b). EG signals and optogenetic pacing of cardiomyocytes optically expressing channelrhodopsin-2 (ChR2) were simultaneously recorded to investigate multimodal feasibility.

Fig. 6: Mesh structured transparent electrode arrays for multimodal sensing and stimulation of neuro-dynamics.
figure 6

a Transparent microelectrode arrays based on an Indium Tin oxide (ITO)/metal grid hybrid (left), metal grid on an ITO film (inset top), metal grid on an ITO island (inset middle), metal grid (inset bottom), and scanning electron microscopy (SEM) image of an ITO island with a metal grid (right). b Langendorff perfusion experiment setup for recording the electrophysiological signal and optogenetics simultaneously on the hearts of mice (left), and the recorded QRS complex with microelectrodes induced by blue light emitting diode (LED) (right). c Single cell in a transparent organic electrochemical transistors (OECTs) array (left), and conformally contacted transparent OECT on a cortical surface of a transgenic rat, stimulated with a blue laser (right). d In vivo evaluation conducted with transparent and non-transparent OECTs, and recorded electrocorticography (ECoG) signals evoked with blue light stimulation depending on excitation intensities. e Illustration of a transparent multichannel Au/PEDOT:PSS bilayer micro-electrode array (MEA). f Microscope image of a bilayer MEA (top), SEM image of a bilayer nanomesh MEA (bottom). g Simultaneous electrophysiological recording (bottom) and two-photon imaging (top right) through an epicortical bilayer MEA on the visual cortex of mice. Panels a and b reprinted with permission from ref. 162, PNAS; panels c and d reprinted with permission from ref. 167, PNAS; panels e, f, and g reprinted with permission from ref. 25, AAAS.

Full size image

To overcome the limitations of passive microelectrode array (MEA), which is poor at recording with a high resolution over a large area, an active MEA can be designed by connecting an amplifier or multiplexer to each passive electrode163,164. The biggest advantage of an active MEA is that it is capable of integratng electrodes with a higher scalability compared to a passive array, and a higher spatial resolution can be obtained when considering the area ratio. Examples of this are organic electrochemical transistors (OECTs), which are composed of arrays of active transistors as channels with organic components, conductive interconnects45,165,166. Figure 6c depicts a transparent and flexible OECT using an Au wire grid and PEDOT:PSS as active channels on a parylene substrate167. By optimally designing the grid width and periodicity of the 3 μm Au mesh interconnection, sufficient optical transparency and electrophysiological measurement fidelity were simultaneously obtained. To investigate the feasibility of low-light artifacts from transparent OECT, ECoG signals from transgenic mice induced by a 473 nm wavelength laser beam were recorded (Fig. 6c, right). This high transparency resulted in a reduction in the reflected light and doubled photo-response intensity compared to the non-transparent OECT (Fig. 6d).

Another way to achieve higher spatial resolution and density is to design smaller size of a single electrode like that of single neuron86,168. Figure 6e shows a 32-channel Au/PEDOT:PSS bilayer MEA based on the nanomesh25. The most challenging part of reducing the site area to the neuron level is the difficulty of maintaining sufficiently low electrochemical impedance for recording. To resolve the challenge, a specialized bilayer structure was implemented by electroplating PEDOT:PSS on the lifted Au nano-mesh structure using a polysterene nanosphere as a mask. By virtue of a double-layer structure based on mesh, electrochemical impedance was 20 times lower than MEAs of graphene or ITO with the same site area and applicable optical transparency. Figure 6f is an optical microscope and SEM image of a single channel microelectrode, shows densely placed voids forming nano-meshes. To observe arousal changes during visual stimulation, two-photon calcium imaging of a single neuron and ECoG were recorded concurrently with the pupil diameter of awakened mice (Fig. 6g). The neuronal action potentials were classified into six level of frequency bands (α, β, γ, high γ, ultrahigh γ, and multiunit). Recorded electrophysiological activities below α band, 8–12 Hz, contains temporal information of Ca2+ indicator (GCaMP6s) imaging, while those of higher frequency bands contain ECoG information. This result highlights the significant benefits of multimodal functions of this work.

Metallic nanowire for neural interface

Another attractive candidate for a transparent neural interface based on metal is to use metallic nanowire (NW). The metallic NW structure has unique properties over the planar metal layer with a mesh structure such as a high effective surface area, an electrical conductivity, and an optical transparency169,170. Owing to these distinctive properties, the nanowire network offers an application capability for optoelectronics. Moreover, the metallic nanowire structure is suitable for interfaces with direct contacts, such as wrinkled tissues, since wire-based construction is much more mechanically flexible than those of carbon or metal oxide.

An optical image of the transparent neural interface mapping ECoG signals in 2D based on a gold nanonetwork (Au–NN) is depicted in Fig. 7a, consisting of randomly patterned Au microelectrodes with 16 channels18. To fabricate Au-NN, chromium, and gold layers were thermally deposited in a photoresistor layer. Next, poly (methyl methacrylate) (PMMA) nanofibers were deposited through electrospinning, serving as a hard mask for nanopatterning. After Au etching with removal of PR, the form of the microelectrodes with Au–NN is completed. One of the key challenges of designing implantable devices with metal NW is wire oxidation, causing a foreign body response that hinders long-term implantation171,172. Owing to the non-reactive property of Au, Au–NN microelectrodes avoid the drawback of wire oxidation. The microelectrode array exhibits 81% of transmittance and low electrochemical impedance of 33.9 kΩ at 1 kHz. In the case of metal electrodes, regardless of their geometrical structures, photoelectric artifacts induced by optical stimulation are often unavoidable. Photoelectric artifacts are proportional to the intensity of the excitation light and the amount of metal surface exposed to light. Thus, the ideal condition to minimize this issue is to limit the exposed portion of the metal surface from the light, maintaining its electrical characteristics. In nano-structured design for the electrodes, the issue of a relatively high electrochemical impedance arises because the conductive path through which electrons can move is somewhat limited compared to the film type173. However, the probability of electron emission by photon excitation is lowered to that extent, so it has advantageous characteristics for integration of optical modalities174. To investigate whether Au–NN is free from photoelectric artifacts for the practical capability of the device, simultaneous in vivo recording and light stimulation (473 nm), which excites ChR2-expressed neurons, was assessed. The device was implanted on the cortical surface of two mice groups: wild type and transgenic mice. Recorded ECoG signals during light stimulation (473 nm) on the cortical surface are presented in Fig. 7b, c. The figure exhibits the eliminated photoelectric artifact contrasted to the Au film electrode, indicating its promising feasibility for the integration of electrophysiology and optogenetics.

Fig. 7: Metallic nanowire-based transparent electrode arrays for multimodal sensing and stimulation of neuro-dynamics.
figure 7

a 16-channel Au-nano network (NN) microelectrodes array for electrocorticography (ECoG) recording. b, c in vivo evaluation of an artifact-free ECoG recording with gold (Au) film (opaque) and Au-NN (transparent) microelectrode arrays with optical stimulation of the cerebral cortex of wild type b and transgenic type (B6 Thy1-ChR2-YFP) mice c. d Schematic illustration of a multifunctional neural interface integrated with a silver (Ag)/Au core-shell nanowire (Ag nanowire (NW)/Au) with poly vinyl alcohol (PVA) gel-coated microelectrodes. e 16-channel AgNW/Au microelectrodes conformally contacted on cerebral cortex including light emitting diode (LED) (left), and the optogenetics in vivo setup of 12-channel AgNW/Au microelectrodes on marmoset brain (right). f 3D graph of the ECoG signal over time, induced by optical stimulation. g Schematic design of transparent ECoG electrode arrays based on AgNW with indium zinc oxide (IZO). h Visualization of neurons with fluorescence microscopy and ECoG recordings through AgNW/IZO microelectrodes of the cortex of mice under urethane anesthesia. Panels a, b, and c reprinted with permission from ref. 18, Wiley; panels d, e, and f reprinted with permission from ref. 22, Wiley; panels g and h reprinted with permission from ref. 179, ACS.

Full size image

To overcome the challenging aspect of chronic implantation, coatings with non-reactive materials can be used175,176. Figure 7d shows one example: a neural interface engaging an Ag/Au core-shell nanowire (AgNW/Au)22. A transparent hydrogel, polyvinyl alcohol (PVA), coated the exterior of the microelectrodes for minimal loss of light, lowering electrochemical impedance, and blocking oxidation. To demonstrate its long-term durability, a reliability test was conducted in saline solution. The Ag/Au electrode with PVA remained electrochemical impedant for 5 months, while the Au-plated electrode without gel remained for 60 days. Plating thin Au on the Ag-NW established higher stretchability and more stable electrical properties compared to pristine AgNW. Ag/Au core-shell NWs performed 83% of transmittance (Fig. 7e), with low electrochemical impedance of 1.1–3.2 Ω sq−1. To investigate the ability of the neural interface to monitor ECoG signals, a viral vector (AAV5-Syn-Chronos-GFP) was injected at a marmoset’s motor cortex and recorded the ECoG evoked from a blue laser (473 nm) (Fig. 7f).

Transparent oxide can be an alternative candidate for defending corrosion of nanowires.39,177,178 Fig. 7g illustrates the schematic design of the transparent electrodes with AgNW and IZO179. IZO not only prevents corrosion of AgNW, but also solves the problem of poor adhesion force, another drawback of AgNW. Multimodal function of optical fluorescence imaging and electrical recording from the rat cortex has been indicated without light loss or photoelectric artifacts (Fig. 7h).

Outlook

To fully understand the precise mechanisms and networks of the brain, monitoring (e.g., two-photon imaging and ECoG) neural activities simultaneously with high spatial and temporal resolution and treating diseases (e.g., optical stimulation and electrical stimulation) with high selectivity are critical. To detect the signal of a specific neuron and stimulate it, the need for a multimodal device, rather than a single tool, is emphasized. Fluorescence imaging, like the two-photon imaging mentioned above, can be highly rated in terms of spatial resolution, but can be hindered by electrophysiological signals in temporal terms. The most important point of such a multimodal device is that the optical signal and the electrical signal must be accurately measured without interfering with each other and work synergistically. This review discussed the necessity of transparent neural interfaces for diagnosing and treating the brain and nervous system in three categories. First, electrophysiological approaches with optical modalities for neural implants were briefly covered, discussing recent studies that integrated two methods. Next, severe challenges from intrinsically non-transparent neural interfaces for simultaneous electrical and optical modalities were introduced, enlightening the necessity of transparent neural implantable devices. Lastly, a detailed presentation of the design of a transparent electrode array with material and structural approaches was posed. Multimodality of transparent neural interfaces can be utilized in various applications, such as the diagnosis and treatment of disorders, opening a broad pathway to next generation neuroscience and medical science.