Techniques for heterogeneous integration of diverse materials from their grown substrates onto receiving substrates of interests in desired layouts have been explored extensively over the past decades1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21. Thanks to the advancement in various micro-assembly techniques, a wide range of proof-of-concept devices and functional systems have been demonstrated, opening up great potential in various applications such as high-resolution displays2,10,22,23,24, flexible optoelectronics25,26,27, bio-integrated electronics28,29, curvilinear electronics30 and many other advanced applications19,31,32,33. These micro-assembly methods are still, however, at their infant stage of development. The lack of mature assembly techniques has hindered the path for commercializing many different devices and applications.

The ability for programmable assembly of tiny chips is of vital importance7,9,13,14,20. One notable merit of such programmable transfer is the effective cost control. The programmable transfer allows a small portion of the components to be transferred at a time, while the remaining devices are still kept on the donor substrate5,14,20. Consequently, any potential risk of device waste can be minimized. Another merit is the defect management. The defective devices can be precluded and only functional devices are selectively transferred to the target substrate. Most importantly, programmable micro-assembly allows the components to be arranged in a format that is different from the original one7,9,22. For example, the spacing and pitch of the transferred objects can be tuned based on an individual user’s need. An example of applications that require programmable assembly is the microscale light-emitting diode (Micro-LED) display7,22,24,29,34,35, which has received intense research interest from the display industry, because of its high brightness, low power consumption, and fast switching speed. For this particular application, millions of Micro-LED chips with size down to a few tens of microns must be prepared densely on source wafers for cost savings and then transfer and printed onto a driven backplane with desired layouts in a relatively sparse form. While these ultrasmall chips are favorable for maximizing the chip density per area, they impose serious challenges for precision assembly. With shrinking the chip size to 100 µm or below, it is reported that Van Der Waals force (VDW) and/or electrostatic force on the chip surface may dominate over the gravitational force36. As a result, the precise and fast release of these tiny devices based on conventional pick-and-place techniques using robotic grippers and vacuum nozzles becomes increasingly difficult. For these reasons, developing alternative techniques for high-throughput programmable assembly of ultrasmall components with a high yield and a fast speed are highly desirable.

Recently, several techniques have been proposed to partially address the programmable transfer capability, but these methods also have their own disadvantages. Among different assembly techniques, stamp transfer printing1,3,14,15,16,18,19 is a popular method that has achieved partial commercial success. This method uses an elastic stamp with engineered surface microstructures1,3,14 to modulate the adhesion strength, making them suitable in transfer printing of tiny, ultrathin, and delicate inks. However, the scalability, printing yield, and placement accuracy of this method remain to be improved further, due to the difficulty in accurate control of the adhesion switchability and stamp deformation37,38. Large area printing commonly suffers from increased defect quantities. Furthermore, the programmable transfer of this approach critically depends on the sophisticated surface engineered microstructures of the stamp, which are usually formed by nanoimprinting or molding techniques14,39,40, resulting in increased cost and manufacturing complexity. Recently, a dry transfer printing method based on atomic diffusion has been proposed to achieve large-area distortion-free printing, but it lacks the ability for programmable assembly6. Laser transfer methods are another type of popular techniques suited for programmable assembly5,7,8,20. For example, laser-induced forward transfer (LIFT)8,41 techniques based on laser ablation of the sacrificial polymer release layer have been developed, but they suffer from placement inaccuracies and surface residuals. Another laser transfer technique relies on localized heating and volume expansion of embedded microspheres in the adhesive layer to modulate the interfacial adhesion and release the inks20. However, volume expansion of the microspheres in the adhesive may cause undesirable position displacement of the inks, imposing practical limitations for high-definition printing. Indeed, the method is only suited for printing components of hundreds of micrometers. In addition, since this method uses raised the temperature to modulate the adhesion, it is not suited for printing some delicate functional micro-objects with poor thermal stability, such as colloidal quantum dots (QDs) and polystyrene microspheres. Recently we have developed a tape-assisted laser transfer technique that is powerful for wafer-level transfer of Micro-LEDs7. However, the relatively high cost of the laser system may be a limiting factor to its use. Fluid-assisted assembly4,10,11 appears to be another high-throughput assembly technique, but the printing accuracy is notably compromised when the chip size shrinks down to 100 µm or below. Furthermore, the chips to be transferred must be placed in a fluid environment, which may not be compatible with the standard micro-fabrication process. Dip-pen-assisted transfer printing with nanoscale printing accuracy has also been reported16,42, but the major limiting factors of this method are the slow transfer speed and the small printing area.

Apart from the requirement of high throughput and programmable transfer, a micro-assembly technique with the capability for assembling materials and devices onto curved substrates is also crucial for developing high-performance curvy electronics7,19,20,26,43. In many cases, foldable or curved devices are highly desirable, which can exhibit expanded functionalities and performances that would otherwise be difficult to achieve from their planar counterparts. Representative examples of such devices include flexible displays2, imagers with curved focal planes44, and smart contact lenses with curvilinear shapes45. However, robust manufacturing technologies for building such devices in large volumes are still under development. Although some printing technologies have emerged to partially address this challenge, they are usually difficult to implement with high efficiency and show poor compatibility with devices with complicated geometries and surface topographies.

In contrast with the conventional stimuli based on pressure46, electrical field47,48, and heat20, recently light stimulus has become an increasingly popular technique for inducing the adhesion switching of photosensitive polymer adhesives, because of its easier temporal and spatial manipulation49,50. Nevertheless, such photosensitive polymers are rarely used for programmable micro-assembly of microscale objects. Herein, we report a rational strategy by synergizing adhesion control and tunable surface topography via patterned light irradiation to achieve large-scale, programmable assembly of various micro-components based on a photo-sensitive polymer. The technique proposed here is ink friendly and compatible with non-planar integration. Large-scale transfer printing is enabled by flood exposure of the photo-sensitive polymer, which results in the polymer switching from the strong adhesion state to the weak adhesion state. Masked UV exposure, on the other hand, provided an additional means to modulate the polymer surface topography. In the latter case, the indented regions of the polymer surface switch to the non-contact state because of the gap formed between the inks and the polymer. With this mask-mediated surface topography, as well as the adhesion switching, micro-objects can be assembled in a programmable format based on users’ needs. To prove these capabilities, diverse ink materials such as indium tin oxide (ITO), gallium nitride (GaN), and gold membranes have been assembled onto various substrates with wafer size up to 4-inch, paving the way toward developing multifunctional electronic systems. Conformal integration of Au nanomembranes, perovskite QDs, and GaN Micro-LEDs onto curvy surfaces have also been demonstrated, highlighting its potential in curvilinear electronics.

Compared to previously reported methods, the technique presented in our research provides a promising, low-cost solution for the large-scale integration of diverse materials for developing advanced electronics with expanded functionalities in the near future. Not only because it offers the capability for large-area (up to 4-inch), programmable assembly of tiny, ultrathin, and delicate components (with a feature size of down to 10 µm, and a thickness down to 250 nm) on demand, but also because the ultraviolet light patterning is a simple, low-cost technique widely used in the modern semiconductor industry such as curing, bonding, and photolithography. This micro-assembly technique is of great potential for exploring next-generation high-performance electronics such as high-resolution displays, bio-integrated health-monitoring electronics, and curvilinear electronics.


Photosensitive polymer with light-regulated adhesion and surface topography

In order to fabricate a photo-sensitive stamp with fast adhesion tunability, we have developed a photo-sensitive polymer adhesive composed of a base acrylic polymer, a trimethylolpropane triacrylate (TMPTA) crosslinking monomer, and a Darocur-1173 photo-initiator. The base acrylic polymer is a mixture of 2-hydroxyethyl acrylate, acrylic acid, and butyl acrylate. The photo-initiator additive endows the polymer adhesive with the photo-responsive property. The crosslinker is added for the purpose of accelerating the decrease in the adhesion, and hardening the polymer. The composition and weight ratio of each component in the polymer is optimized to maximize the photo-responsive time. The detailed procedure of preparing the photo-sensitive polymer can be found in the method section. The polymer adhesive is then coated onto a thin polyethylene terephthalate (PET) or polyolefin (PO) sheet by an automatic film applicator to form a polymeric adhesive tape (Fig. 1a), which is also termed as the photo-sensitive stamp in the following. A cross-sectional scanning electron microscope (SEM) image of the photosensitive tape is given in Supplementary Fig. 1. Initially, the stamp possesses very a strong adhesion of 132.2 N cm−2 based on the peeling test (Supplementary Fig. 2) and remains a relatively flat topography (Ra = 7.74 nm) as evidenced by the atomic force microscope (AFM) image shown in Fig. 1a. Upon 365 nm ultraviolet flood irradiation with a power density of 20 mW cm−2 for only 3 s, crosslinking and photochemical reactions of the polymer rapidly occur, leading to a fast drop in the adhesion strength to nearly 1.1 N cm−2 and an increase in surface roughness (Ra = 11.3 nm) of the adhesive layer as evidenced by the AFM image shown in Fig. 1b. Figure 1c shows the dependence of the adhesion strength as a function of the exposure time. It is observed that the adhesion strength decreases from 132.2 N cm−2 to 1.1 N cm−2 with increasing the exposure time from 0 s to 3 s. The corresponding switchability (defined by the ratio of the adhesion strength after exposure to the adhesion strength before exposure) as a function of the exposure time is shown in Fig. 1d. Depending on the irradiation time, the adhesion switchability can be as large as 117.5, indicating the great possibility of retrieving and delivering the printed inks using this polymer. We notice that the adhesion reduction of our polymer adhesive against the UV exposure time is much faster than commercially available UV tapes (Supplementary Fig. 3). In the latter case, it will take a couple of minutes to decrease the adhesion to a comparable value. The underlying mechanism of this light-induced adhesion reduction can be summarized as two factors. One of them is the formation of a cross-linked network structure which is harder and less sticker when exposed to ultraviolet stimuli51. The other factor owes to the increased roughness of the polymer surface upon ultraviolet exposure, which leads to the formation of a smaller effective contact area at the interface between the adhesive and the object surface, thereby reducing the interfacial adhesion.

Fig. 1: Characteristics of the photosensitive polymer with switchable adhesion.
figure 1

a Schematic of the initial strong adhesion state of the polymer without light irradiation (Top), and AFM surface morphology of the adhesive, showing a relatively low surface roughness. b Schematic of the weak adhesion state due to the light-induced solidification (top), and AFM image of the polymer surface after UV light irradiation, showing increased surface roughness. c Adhesion strength of the photosensitive polymer as a function of the UV exposure time. d Adhesion switchability of the polymer as a function of the exposure time.

The aforementioned results suggest the polymer surface adhesion is closely related to its topography and the degree of polymer crosslinking. Keeping this in mind, we have further studied the change of polymer surface adhesion and topography after ultraviolet irradiation through an optical mask, as schematically shown in Fig. 2a. Similarly, a 365-nm ultraviolet stimulus at 20 mW cm−2 for a duration of 1.2 s is applied to the polymer stamp placed on a glass support via a photo mask with circular patterns in different diameters. It is interesting to note that the mask patterns have been well duplicated to the polymer surface, result in forming a sharp boundary separating the exposure region and the masked region (Fig. 2b). Laser confocal scanning microscope analysis further indicates that the exposure region becomes higher than the non-irradiation region (Fig. 2c), indicating of the 1.55 µm polymer growth in the irradiation region. It should be noted that the glass support holder below the polymer stamp can effectively prevent oxygen in the air from attacking the polymer/glass interface during UV irradiation, which is essential for promoting the polymer growth effect. It is noted that the edge height of the exposure area is higher than in the center (Fig. 2d). Further investigations indicate the height difference is dependent on the UV irradiation time. With increasing the exposure time from 0.8 to 2.5 s, the polymer height gradually expands from 0.7 to 2.4 µm. Further increasing the exposure time to 3 s, however, does not increase the polymer thickness significantly (Fig. 2e). This implies the polymer growth effect tends to reach its threshold. Apart from the modulated polymer thickness, we also confirm the adhesion strength is remarkably lower in the irradiation region than the masked region. For instance, with a 2.5 s UV stimulus, the adhesion strength of the photosensitive polymer in the masked region is around 132 N cm−2, whereas it reduces to 12.7 N cm−2 in the exposure area, as shown in Fig. 2f. These results imply that it is easy to produce a smart adhesive stamp with both the adhesion and the polymer height modulated at different locations via simple masked light exposure. It should be noted that the polymer growth effect is not observed in the commercial UV tape that we have tested, indicating this effect is highly associated with the monomers used in the polymer synthesis.

Fig. 2: Mask-defined adhesion switching and polymer growth of the photosensitive polymer.
figure 2

a Schematic of the mask-defined irradiation applied to the light-sensitive polymer. b Optical image of the photosensitive polymer after 1.2 s UV irradiation using a photomask, showing the increased polymer thickness at the exposure area. c Laser confocal 3D image of the polymer surface corresponding to the area marked by the red square in (b). d Cross-sectional profile of the polymer surface under 1.2 s UV irradiation, scanning along the yellow arrow marked in (c). e polymer height as a function of the UV exposure time. f Measured adhesion strengths at the exposure area, and masked area. The exposure time is 2.5 s. g Schematic of the light-induced polymer growth.

The observed polymer growth can be attributed to the polymerization-induced diffusion52,53, as schematically shown in Fig. 2g. Upon UV light irradiation, the crosslinking monomer TMPTA concentration is decreased in the exposure area because of the partial photo-polymerization. On the other hand, the monomer concentration in the masked region is higher initially. As a consequence, the concentration difference (or gradient) of the TMPTA monomer between the dark region and lighting region, drives the monomer to diffuse from the dark region to the exposure region. With increasing the light exposure time, the continuous diffusion of the monomer leads to increased mass transport (or polymer thickness) in the exposed area. The fact that the polymer height is higher at the edge of the exposure area then in the center further indicates crosslinking monomers are preferentially accumulated and polymerized at the edge, and then diffused toward the center of the exposure area. A detailed static analysis based on the diffusion model is given in the supporting information and Supplementary Fig. 4.

Wafer-scale transfer printing

Inspired by the unique adhesion tunability of the photo-sensitive polymer, we have developed a versatile route for achieving high-throughput, wafer-level assembly of components with different dimensions, as schematically shown in Fig. 3a. The process begins with preparing printable device components, which are weakly bonded to a donor substrate (Fig. 3a-i). A photo-sensitive adhesive (or stamp in the following) is then imprinted onto and forms conformal contact with the printable devices on the donor substrate (Fig. 3a-ii). By peeling off the adhesive stamp (Fig. 3a-iii), wafer-level device components can be readily retrieved by the stamp (Fig. 3a-iv). In order to further transfer the components from the photo-sensitive stamp to another receiver, ultraviolet light is flooded over the photo-sensitive adhesive surface (Fig. 3a-v). By imprinting the photosensitive stamp with retrieved inks onto the receiver (Fig. 3a-vi), and peeling off the adhesive stamp (Fig. 3a-vii), these inks can be reliably released to the target substrate coated with an adhesive layer (Fig. 3a-viii).

Fig. 3: Wafer-scale transfer printing.
figure 3

a Proposed wafer-scale micro-assembly procedure based on the photosensitive adhesive. b Optical images of the original gold nanomembrane inks on a 2-inch silicon (i), the gold nanomembranes picked up by the first photosensitive polymer tape (ii), the gold nanomembranes transferred to the second photosensitive tape (iii), and the gold nanomembranes transferred to a glass substrate (iv). c Optical images of gold nanomembrane inks transferred onto a PO substrate. The feature size of the gold nanomembrane is ranging from 10 to 30 µm. The printing area of all gold nanomembrane arrays are <1 cm × 1 cm. d Photograph images of the large-area gold nanomembranes transferred to a 2-inch glass substrate. e Photograph images of the large-area gold nanomembranes transferred to a 4-inch PET substrate. For all transfer printing tests shown in (be), flood exposure time is 2.5 s.

The working principle behind this method is interfacial adhesion manipulation through UV light irradiation, as schematically shown in Supplementary Fig. 5. Initially, the photo-sensitive polymer has large adhesion strength, such that the interfacial adhesion between the photosensitive polymer adhesive and the device \(F_{\rm{polymer} - \rm{device}}\) is much higher than between the device and the donor substrate \(F_{\rm{device} - \rm{donor}}\). In other words, the following inequity is satisfied, and therefore the device can be picked up by the photosensitive polymer (Supplementary Fig. 5a)

$$F_{\rm{polymer} - \rm{device}} \,>\, F_{\rm{device} - \rm{donor}}$$

Upon the UV light irradiation, \(F_{\rm{polymer} - \rm{device}}\) becomes notably reduced due to the polymerization of the photo-sensitive polymer adhesive, and it becomes much smaller than the interfacial adhesion between the inks and the receiving substrate coated with an assisted adhesion layer \((F_{\rm{device} - \rm{receiver}})\). Thus, the device on the polymer adhesive can be released to the adhesive receiver (Supplementary Fig. 5b). That is, ink release is given by the condition

$$F_{\rm{polymer} - \rm{device}} \,< \, F_{\rm{device} - \rm{receiver}}$$

To demonstrate the extraordinary ability of the designed photosensitive stamp in transfer printing of tiny, ultrathin, and delicate components, a wafer-level 250 nm-thick transferable Au membrane with different dimensions ranging from sub-millimeter to millimeter is fabricated on a 2-inch Si/SiO2 substrate (Fig. 3b-i). The detailed Au membrane fabrication procedure can be found in Supplementary Fig. 6. As shown in Fig. 3b, various Au patterns with feature sizes ranging from 500 µm to 5 mm are completely picked up using the photosensitive stamp and well reproduced from their original ones, without any noticeable distortion of the shape and size (Fig. 3b-ii). Furthermore, thanks to the high initial adhesion of the photo-sensitive adhesive, the wafer-level acquisition of these Au patterns to the stamp can be finished in a few tens of seconds, implying the high efficiency of this process (Supplementary Video 1). Favorably, since the adhesion strength can be simply reduced by external ultraviolet stimuli, these devices can be flipped over and transferred to a fresh adhesive tape under optimized UV irradiation duration (Fig. 3b-iii, and Supplementary Video 2). In this work, we find the optimized flood UV exposure time is around 2.5 s based on preliminary transfer printing tests using small gold membrane samples (Supplementary Fig. 7 and Supplementary notes). Repeating these steps enables the possibility of multiple transfers of the objects among various substrates. For the purpose of illustration, the same gold membrane patterns can be further completely transferred to a 2-inch glass substrate coated with a SU8 adhesive layer (Fig. 3b-iv), while without sacrificing the printing yield obviously. While only the assembly of gold nanomembranes has been demonstrated so far, this process is actually also compatible with transfer printing diverse ink materials such as ITO, GaN and AZ4620 photoresist (Supplementary Fig. 8).

Notably, this method can be easily extended to assemble wafer-level discrete objects with smaller feature sizes below 10 µm. To demonstrate this capability, high-density, 250 nm thick Au arrays are fabricated on three typical silicon wafers of different dimensions (<1 cm by 1 cm, 2 inch, and 4 inch), followed by transfer printing to different substrates using the photo-sensitive stamps. Regardless of the substrate size and substrate material, in all cases, high-density Au arrays are completely picked up using the adhesive stamp and successfully delivered onto the receiving substrate, as shown in Fig. 3c–e. Figure 3c shows the high-quality Au patterns with feature sizes ranging from 10 to 30 µm printed onto a PO substrate. The printing area for each Au array is <1 cm × 1 cm. Regardless of the feature size (10–30 µm) and spacing (25–43 µm), the printing yield is nearly 100% for all Au arrays across the substrate. In a most recent effort, high-fidelity printing of 40 nm-thick, 4 µm-sized Au array on a 7.5 µm pitch has been further demonstrated (Supplementary Fig. 9), highlighting the potential to further shrink the printing feature size. Similar strategies are used to assemble large-area Au nanomembranes to larger wafers, including a 2-inch glass wafer (Fig. 3d) and a 4-inch PET substrate (Fig. 3e, and Supplementary Fig. 10). Although the printing areas are substantially increased, the transfer yields are not obviously compromised in both cases. For example, a high printing yield of more than 99.9% can be achieved for the 4-inch transfer printing. While not further explored in this work, we believe our technique is compatible with 6-inch or even bigger large area printing onto diverse substrates. In contrast with existing technologies which often limit to small-area transfer, our technique has the distinct potential for large-scale assembly of various components.

It should be noted that, for the above transfer printing tests, all receiving substrates (except for the UV adhesive tape) have been coated with a supplementary 2 µm-thick SU8 adhesive layer (Microchem SU8 2002). This adhesion layer not only helps to compensate for the residual weak adhesion of the photosensitive stamp after exposure, but also helps to overcome the non-neglectable surface electrostatic force and/or VDW force associated with the tiny inks. It was reported that these surface forces are dominant factors affecting the reliable release of the tiny inks if they are not properly eliminated.

Apart from the wafer-scale transfer capability, placement accuracy and printing uniformity of these transferred microscale objects are also critical factors to be addressed for subsequent processes. Examples of such post-processes include the metallization to form ohmic contacts on the transferred device, and the via opening for multiple device interconnection. All these processes raise the strict requirement for printing uniformity and minimal registration error. Large position offsets can lead to disastrous results, for instance forming open circuits due to misalignment. Printing uniformities at five points across the wafer have been checked (Supplementary Fig. 11). The consistent printing yields and feature sizes at these different locations indicate the distortion of the transferred pattern is minimized. We further quantitatively evaluated the transfer accuracy of our transfer method, as shown in Supplementary Fig. 12. We notice that the printing offset along both the X-axis and Y-axis shows minimal correlation with the printing position. Regardless of the transfer area, a statistical analysis of the data among five different areas across the 4-inch wafer reveals that the biggest printing offset is only ±0.4 µm. The high printing accuracy across the large printing area is another merit of the current technology.

Programmable transfer printing

While wafer-level transfer of all objects from the donor substrate to the receiver in a 1:1 manner is popular, there are many cases that require the selective transfer of a portion of components from the designated area. Herein we demonstrate our technique is feasible for programmable transfer. Although laser transfer and PDMS stamp transfer1,3,15,41 techniques can partially address this challenge, we believe photo-triggered selective transfer is remarkably simpler to implement and more cost-effective, since the stamp here is simply formed via mask-defined light pattern, which doesn’t involve the use of expensive facilities and complicated lithography process.

To illustrate this extraordinary capability, we use a similar transfer process mentioned earlier in Fig. 3a, but after the flood exposure of the fifth step, the photo-sensitive polymer stamp is further selectively irradiated using a photomask (Fig. 4a). The initial flood exposure decreases the interfacial adhesion of the stamp, while the subsequent mask exposure triggers the photo-induced polymer growth in the exposure area, resulting in forming corrugated surface topography. Using this modified process (Fig. 4b), we have successfully transferred discrete GaN inks to a receiver substrate in a programmable manner. The fabrication details of the GaN inks can be found in the method section. An example of two opposite patterns consisting of GaN inks assembled in a heart shape is shown in Fig. 4c, d, for which a 1.5 s flood exposure is used, followed by a 1.5 s masked exposure. The former pattern shown in Fig. 4c corresponds to the GaN ink array transferred to the final PET substrate, whereas the latter one in Fig. 4d is the GaN ink array remaining on the adhesive stamp. With this dual exposure strategy, these patterns are well reproduced from the respective photo mask. It should be noted that both the flood exposure time and the mask exposure time here must be optimized to achieve a high transfer yield. If the initial flood exposure is too long, the subsequent mask exposure will not play a critical role in inducing the polymer growth because of the full solidification of the UV-sensitive polymer. On the other hand, if the initial flood exposure is too short and the masked exposure is too long, it will be difficult for reliable release of the retrieved inks. Indeed, there is a clear evidence that more defective inks will be generated if the initial flood exposure is missing (Supplementary Fig. 13). We attributed this to the large initial adhesion of the stamp, which is hard to be compensated by the adhesive layer on the receiver if the flood exposure is not used.

Fig. 4: Programmable transfer printing and its mechanism.
figure 4

a Heart-shaped mask used for selective UV irradiation. b Optical image of the GaN ink array on the photosensitive polymer adhesive before selective exposure. The heart-shaped shadow shown in (b) is a guide to the eyes, and corresponds to the region where will be not exposed. c Optical image of the heart-shaped GaN ink array transferred to the receiver by 1.5 s flood exposure and 1.5 s selective exposure. d Optical image of the GaN inks left on the original adhesive tape after programmable transfer. e Confocal 3D topography of the GaN ink array and the adhesive surface after mask exposure but before the transfer, revealing the GaN Inks in the masked area are higher, whereas the GaN inks in the exposure area are embedded in the polymer layer. The scanning area is taken from the region marked by the red rectangle shown in (b). f Height profile of the GaN devices scanning along the yellow line marked in (e). g Schematical illustration of the working principle of programmable transfer. A gap is formed between the irradiation area and masked area, due to the light-induced polymer redistribution. GaN Inks in the exposure area are completely sucked in the irradiated polymer. This is the major reason why the GaN inks in the masked region are transferred to the final substrate.

To gain a deeper understanding of the selective transfer mechanism, we have examined the 3D profile of the inks in the masked region and non-masked region immediately after mask exposure, but before transfer (Fig. 4e). The scanning area in Fig. 4e is taken from the corresponding area marked by the red rectangle in Fig. 4b. We find that the polymer thickness in the exposed area is higher than in the masked region, consistent with the experimental observation earlier described in Fig. 2. However, all GaN inks in the irradiation area are embedded in the crosslinked polymer (Fig. 4e). On the other hand, GaN inks in the masked region still maintain their original states, and no such embedding effects are observed (Fig. 4e). Consequently, GaN inks in the masked region, corresponding to the area where the inks will be transferred (Fig. 4c), are higher than those in the irradiated region. The actual height difference is around 1.6 µm in this context (Fig. 4f).

From the above analysis, we conclude that the programmable transfer mechanism can be attributed to the light-regulated height difference of the inks between the exposure area and the masked area, as well as the reduced interfacial adhesion strength of the stamp, as schematically shown in Fig. 4g. The masked ultraviolet exposure (Fig. 4g–i) creates complicated mass transportation of the polymer from the masked area to the irradiated area, resulting in forming a locally indented region and elevated region on the stamp (Fig. 4g-ii). Inks in the elevated region will be contacted with the receiver, while those in the contracted region will not (Fig. 4g-iii). Because the receiver is stickier than the stamp (after the first flood exposure), this facilitates the selective assembly of the elevated inks onto the receiver (Fig. 4g-iv).

Curvy surface-compatible transfer printing

Most of the devices and functional systems are built on planar substrates. However, many emerging applications have driven the development of curvy electronics. Examples of such applications include contact lens displays45, and electronic eyes with curved focus planes44. These devices are generally more difficult to build, since conventional lithography is mainly based on planar substrates. Here we show our technique is well compatible with the micro-assembly on curved surfaces. This arises from the soft nature of the thin sheet adhesive stamp which allows conformal contact with the curvy surface, as schematically shown in Fig. 5a. To demonstrate this capability, micro-scale ultra-thin Au membranes with dimensions of 40 µm by 70 µm have been successfully delivered onto a curvy glass bottle surface with a high transferring yield, as shown in Fig. 5b. It is observed that these Au membranes are free from shape distortions and wrinkles, revealing the high fidelity of this method. We further compare the electrical property of an individual Au pixel before and after transfer and no obvious change of the resistance is found, as indicated by the I-V curves in Fig. 5c.

Fig. 5: Transfer printing on curved surfaces.
figure 5

a Schematical transfer printing process on curved surfaces. b Optical and SEM images of the gold nanomembranes transferred to the curvy bottle surface, with a flood exposure time of 1.2 s. c I-V characteristics of the gold nanomembrane. d Schematic of the PQD inks formed on a glass substrate by sequential drop coating (Left), and images of the PQD inks before excitation (top right), and after excitation (bottom right). e Optical images of the PQD inks transferred to the bottle surface before excitation (left), and after excitation (right). The flood exposure time is 2.5 s. f Photoluminescent (PL) characteristics of the PQDs before transfer and after transfer.

In addition to Au pattern assembly, QD transfer printing35 is another potential application of the current technique, which can be even more challenging. Perovskite QDs (PQDs) are a class of nanocrystals that have the potential use for high-performance QD displays. However, it is commonly challenging to assemble PQDs onto a foreign substrate based on existing methods since they are not very stable and inevitably suffer from the performance degradation when exposed to water and oxygen-rich environment. Fortunately, based on our dry transfer technique, such PQDs blended with polystyrene (PS) are readily printed onto a glass vessel surface with high fidelity. Details of the PQD synthesis and characterization can be found in Supplementary Fig. 14. For purposes of illustration, in this particular case, the PQD dot pattern is formed on the glass substrate by simple sequential drop casting a few PQD droplets (Fig. 5d), and then transferred printed to the sidewall of the glass bottle (Fig. 5e) using the photo-sensitive polymer. The fact that the photoluminescent intensity of the PQDs after the transfer is comparable to those before transfer (Fig. 5f), implying no obvious degradation of the QD quality arising from the printing. This can be further verified by the observation that no extra emission can be detected from the original glass substrate both visually and optically.

Micro-LED transfer printing for displays

As a proof of concept, we further demonstrate the use of our technique for assembling Micro-LEDs for display purposes. Micro-LEDs are miniaturized LEDs with typical dimensions of a few tens of microns, which have received intense research interest for developing high-resolution displays, because of their excellent characteristics such as short response time, and low power consumption, high brightness, high stability, and long lifetime7,22,24,34. However, the construction of a high-definition Micro-LED display screen is facing a large number of technical problems at this stage, such as mass transfer, full colorization, and defect management. Among these issues, a highly reliable approach to assembling a large quantity of Micro-LEDs in an ultra-precise and low-cost manner is critically important for the industrialization of Micro-LED display technology.

Here, by exploiting the adhesion tuneability of the photo-sensitive polymer adhesive upon UV stimuli, we show that dense Micro-LEDs can be readily assembled onto the target substrate in a simple manner. For the purpose of illustration, standard micro-fabrication and wet etching techniques are used to fabricate printable Micro-LEDs on commercially available Si-based LED wafers (Supplementary Fig. 15) in a configuration such that each Micro-LED is suspended and tethered by SiO2 using a similar strategy reported by Kim et al.54 (Fig. 6a). The detailed fabrication procedure of such transferrable GaN Micro-LEDs can be found in Supplementary Fig. 16. Following the transfer printing procedure shown in Supplementary Fig. 17, these weakly bonded Micro-LED devices on Si <111> can then be readily released by breaking the tethers and then transferred to a target transparent glass coated with a SU8 as the adhesion layer (Fig. 6b). Thanks to the robust transfer process and large adhesion switchability of the photo-sensitive stamp, nearly 100% transfer yield of the Micro-LED has been achieved. SEM inspection and energy-dispersive X-ray spectroscopy (EDS) analysis confirm the transferred chips are free from contamination (Supplementary Fig. 18). Minimized electrical degradation is confirmed by comparing the I-V characteristic of Micro-LEDs on the original growth Si <111> wafer with those transferred to the glass substrate, indicating the transfer process has minimized side-effect on the device performance (Fig. 6c). However, the emission wavelength is slightly redshifted, owning to the strain release during undercutting the LED devices55 and/or junction heating56 (Fig. 6d).

Fig. 6: Micro-LED transfer printing.
figure 6

a Optical images of tethered Micro-LEDs formed on a silicon substrate before transfer printing. b Optical images of Micro-LED arrays transfer printed to a glass substrate under different magnifications. Flood exposure time is 2.5 s. c I-V characteristics of the Micro-LED before transfer and after transfer. d EL spectra of the Micro-LED before transfer and after transfer. The injection current is 6 mA. e Optical image of the Micro-LEDs left on the original adhesive after selective transfer. f Optical image of the arrow-shaped Micro-LEDs selectively transferred onto a PET substrate, with a 1.5 s flood exposure and a 1.5 s masked exposure. g Emission image of the printed arrow-shaped Micro-LED array with one pixel switched on by probes. h Optical images of selectively printed arrow-shaped Micro-LEDs integrated onto the curvy surface of a glass bottle (Left, and top right), and the same array with one pixel switched on, after integration onto the curvy surface (bottom right). i I-V characteristics of the Micro-LED on a flat substrate and a curvy substrate, j EL spectra of the Micro-LED on a flat substrate and a curvy substrate.

Apart from the above one-off transfer (i.e., non-selective, full transfer), Micro-LEDs can be also assembled into programmable formats based on user needs through flood exposure followed by selective UV irradiation using a photomask. An example of the programmable Micro-LED pattern arranged in “arrow” shape can be found in Fig. 6e, f. Figure 6e shows the Micro-LED pattern remaining on the original tape, whereas Fig. 6f shows the Micro-LEDs transferred to the receiving substrate. Figure 6g shows the emission image of one pixel switched on in the printed array. Clearly, one transfer results in forming two opposite patterns, including a negative pattern and a positive pattern. These selectively printed Micro-LEDs can be further conformally pasted and integrated onto curved bottle surfaces, highlighting the potential in developing curvy electronics, as shown in Fig. 6h. It is clearly shown that the arrow-shaped Micro-LEDs on the curved surface (with a radius of curvature of 20 mm) are still fully functional, without apparent electrical (Fig. 6i) and optical degradation (Fig. 6j). By simply modifying the mask patterns, various programmable patterns can be achieved (Supplementary Fig. 19), highlighting the extraordinary capabilities of the programmable transfer printing of our proposed technique based on light-induced adhesion switching.

These selectively printed discrete Micro-LEDs, followed by metal wiring, can be further built into Micro-LED display devices. Figure 7a schematically shows one exemplar device architecture, which consists of the glass substrate, Micro-LEDs, SU8 insulating layers, as well as the metal interconnects. The optical image of the resulting device consisting of 25×20 parallelly addressed Micro-LEDs is shown in Fig. 7b. The fabrication details of this device can be found in Supplementary Fig. 20. Figure 7c, d shows the emission images of the device with an injection current of 1 mA, taken under the bright field and the dark field, respectively. It is observed that the emission across the whole array is highly uniform, and all pixels are fully functional. The I-V characteristic of the Micro-LED display device is shown in Fig. 7e. It is observed that the device shows normal rectification behavior similar to conventional large-size LEDs, with a typical turn-on voltage of around 2.5 V. Figure 7f shows the EL spectra as a function of the injection current. The peak emission wavelength is almost constant with increasing the injection current from 0 to 50 mA, revealing of the working stability of this Micro-LED display device.

Fig. 7: Interconnection of transfer-printed Micro-LEDs for displays.
figure 7

a Exploded view of the interconnected Micro-LED display device architecture. b Optical image of the resulting 25 × 20 Micro-LED array. c Bright-field image of the full array switched on. d Dark-field image of the full array switched on. e I-V of the interconnected Micro-LED display device. f Current-dependent EL spectra of the interconnected Micro-LED array device.


In summary, we have presented a simple, yet powerful transfer printing technique based on a light-switchable polymer adhesive, which allows selective, large-area transfer printing of various ink materials onto both planar and curvy substrates. Our technique overcomes some limits of currently available transfer printing techniques. First, the stamp adhesion switchability is simply tuned by external ultraviolet irradiation which is widely used in the modern semiconductor industry such as curing, bonding, and photolithography. The ultraviolet stimulus has minimized damage to inks to be transferred since it can be remotely delivered to the target object. Second, the current method facilitates the formation of a patterned stamp through simple masked irradiation without the need for accessing expensive lithography and nano-imprinting tools, which would be otherwise required in other techniques. Most importantly, the current technology has the ability for fast, large-area assembly (up to 4-inch) of ultrasmall, and delicate functional components in a programmable manner into spatially organized arrangements with arbitrary layouts. A more detailed comparison of our technique with the existing transfer printing methods is summarized in Supplementary Table 1, which clearly reveals some distinct advantages of our technique, including the ultrathin ink processing capability, higher printing accuracy, and larger printing areas. The proposed transfer printing technique based on the light-switchable polymer adhesive here can open up a wide range of potential applications such as high-resolution displays, and flexible electronics. We notice the placement accuracy of the current method is poorer than the dip-pen assisted transfer technique42, but the transfer area per time and transfer speed are significantly improved.

We note that some other tapes such as thermal release tape (TRT)57, and water-dissolvable tapes58 have been used for transfer printing. These methods, however, are commonly only suited for one-off transfer, lacking the capability for selective transfer. One exception is the use of a pulsed laser in combination with a TRT tape to achieve selective transfer20, but the process is more complex, and more expensive. We also address one limitation of the current method. Because of the non-reversable adhesion switching of the polymer adhesive, the photosensitive tape cannot be repeatedly used for transfer printing. Nevertheless, the extremely low manufacturing cost and the superior capability for large-area printing of the photosensitive polymer, make it still a good choice for large-scale, programmable assembly of micro-scale inks. Furthermore, we notice some new light-responsive polymer adhesives49,50 with repeatable, reversible adhesion switching have been reported recently. However, whether these polymers are feasible for large-scale programmable assembly of micro-scale objects or not, is an open question that remains to be clarified. The development of such photosensitive, adhesion-switchable adhesives may represent the future research direction of extending the current light-regulated micro-assembly technique.


Light-switchable polymer synthesis

butyl acrylate (BA), 2-hydroxyethyl acrylate (HEA), and acrylic acid (AA) monomers with a mass ratio of 82.7:10:7.3 were mixed uniformly for 30 minutes to form an acrylic prepolymer. The obtained mixture was then further mixed with trimethylolpropane triacrylate (TMPTA) crosslinking agent (0.3 wt%) and 1173 photoinitiator (1 wt%) for 2 h to form the photosensitive adhesive.

Adhesion strength tests

A 1 cm × 5 cm photosensitive stamp was attached onto a cleaned glass slide. Then, it was connected to the load cell of the peeling apparatus (Supplementary Fig. 2). The 90° peeling force that was used to quantify the interfacial adhesion strength of the photo-sensitive stamp was measured at various ultraviolet intensities and exposure time. The peeling force was calculated by averaging the maximum peeling force from three repetitive trials. Adhesion strength was then defined as the measured peeling force divided by the polymer adhesive area.

Fabrication of ultra-thin transferable Au membranes

The fabrication procedure of ultra-thin transferable Au membranes is schematically shown in Supplementary Fig. 6. A SiO2 etch sacrificial layer was deposited on a clean Si <100> wafer using a plasma enhanced chemical vapor deposition (PECVD) system (Oxford Plasmalab 800Plus). On top of the SiO2 film, an AZ2035 photoresist layer was then spun coated, and patterned by optical lithography. The next step was to deposit 250 nm, Au, using a Metal E-beam Evaporator (DE400) at a base pressure of <5 × 10−7 Torr, followed by a metal lift-off process using acetone solvent to strip off the photoresist. The sacrificial SiO2 layer was then undercut by diluted BOE etch for 60 s using the patterned Au as a mask, resulting in form of suspended Au membrane arrays weakly bonded to the substrate and ready for printing.

Fabrication of transferable GaN Micro-LEDs

Si-GaN Micro-LEDs were fabricated from commercial GaN/Si <111> wafers, with epi-stacks consisting of 150 nm p-GaN dopped with Mg, 200-nm InGaN/GaN Multiple Quantum Wells (MQWs), 1600-nm n-GaN dopped with Si and 1450-nm GaN buffer (Supplementary Fig. 15). The fabrication procedure is schematically shown in Supplementary Fig. 16. The process began with depositing a transparent conducting film ITO (250 nm) on p-GaN by using a sputter system, followed by rapid thermal annealing at 550 °C for 5 min in O2 ambient to form an ohmic contact to p-GaN. A wet etching using HCl solution was conducted to form well-defined ITO pattern using photolithographic defined AZ4620 resist as a mask. The same photo-resist AZ4620 mask was then used to expose the regions of n-GaN by dry etching using an inductively coupled plasma (ICP) system (Oxford Plasmalab system 133). After stripping off the residual resist, A 500 nm layer of SiO2 was deposited by PECVD, followed by BOE wet etching to form a patterned SiO2 mask using lithography defined AZ 4620 photoresist. This SiO2 mask was then used to define isolated GaN mesa array structures, which also define the emission area (50 μm × 80 μm). A 300 nm SiO2 passivation layer was deposited by PECVD and patterned by photo-lithography. Both p-contact and n-contact metal pads (Ti:10 nm/Au:350 nm) are deposited via e-beam evaporator, followed by a lift-off process using a patterned AZ2035 photoresist. The formed contact pads were then annealed at 450 °C for 45 s in O2 ambient. Finally, anisotropic undercut etching of the silicon was performed by immersing the sample in a solution of TMAH at 85 °C for 30 min to form tethered Micro-LED structures for transfer printing.

Fabrication of GaN inks

GaN rectangular mesa structures were formed by lithography and ICP deep dry etching down to the Si substrate, based on a GaN wafer with a 4 μm GaN layer on Si <111>. The formed GaN mesas were then served as masks to undercut the silicon underneath, results in forming printable GaN inks weakly bonded to the substrate.

Transfer printing using a photo-sensitive polymer stamp

The photosensitive polymer adhesive was coated onto either a PET or PO backing layer by an automatic thin-film applicator to form a photo-responsive stamp. The polymer stamp was then imprinted onto inks on the source wafer by using a roller. By Peeling off the polymer stamp, inks can be retrieved from the source wafer. The stamp together with the inks, was then imprinted onto the receiver coated with a SU8 adhesive layer, which helps to release the inks from the stamp. Afterwards, UV light exposure was conducted to reduce the adhesion of the stamp. Finally, ink transfer printing onto the receiver was enabled by removing the photosensitive stamp.

Interconnect of printed Micro-LEDs for displays

The Micro-LED devices were transfer printed from the source wafer to the target substrate, using the procedure described elsewhere (Supplementary Fig. 17). Photolithography using a negative-tone resist SU8-2002 as sidewall passivation layer was performed, such that both p-pad and n-pad are exposed. After SU8 curing (ultraviolet exposure 60 s and baking at 100 °C for 20 min in air), metal interconnects (Ti/Al/Ni/Au in desired thickness) were deposited by a sputter system (Kurt. J. Lesker), followed by AZ 2035 photoresist patterning and metal lift-off in acetone.

Materials and device characterization

The analysis of morphology and structure was conducted with Metallographic microscope (Leica), Laser scanning confocal microscope (Olympus), Scanning electron microscope (S-4800 FESEM), Atomic Force Microscopy (Dimension Edge, Bruker), and Profile-system (Dektak XT, Bruker). The material was characterized with X-Ray Diffraction (Smartlab 9kw, RIGAKU). Electrical measurements were performed with a semiconductor parameter analyzer (B1505A, Keysight) combined with a probe station (TS2000-HP, MPI). Optical measurements of the emission spectra were performed with a probe station (Prober PG 2101, Detech).