Flexible microLEDs without compromise

Photonic die transfer preserves the performance of InGaN microLEDs while enabling flexible devices to deliver brighter and more efficient light output.

BY PRANAV GAVIRNENI AND WILLIAM WONG FROM THE UNIVERSITY OF WATERLOO AND HARRY CHOU AND VIKRAM TURKANI FROM PULSEFORGE

Medical professionals use numerous treatments to help their patients, including quite a few that involve light. There’s photodynamic therapy, using light to activate medicines that kill abnormal, pre-cancerous, or cancerous cells; UV phototherapy, which helps clear up stubborn skin conditions, such as psoriasis, eczema, and vitiligo; and red-light therapy, a treatment for improving blood flow, reducing joint pain, and easing inflammation.

When light plays a role in a patient’s treatment, it’s ideally applied directly to a specific area. And promising to do just that with exceptional efficacy are flexible biomedical systems, which can deliver a new level of control over how light interacts with the human body. Such systems could be based on conformal devices, which wrap onto soft tissue to deliver highly localised optical stimulation, enabling functions not possible with rigid platforms. However, realising this capability requires more than just mechanical flexibility – it demands light sources that remain efficient, stable, and precisely controlled while operating on deformable surfaces.

Figure 1. A microLED emitting blue light during operation, illustrating the high brightness and small size that enable dense arrays with fine spatial control.

Meeting these performance demands are microLEDs. They have a high brightness, fast response, and micron-scale dimensions, making them well suited to dense arrays, where they deliver light with high spatial resolution (see Figure 1). It’s a combination of attributes that’s essential for applications where optical output must be confined and controlled within a small form factor, such as wearable and implantable systems.

Unfortunately, when microLEDs are considered for these applications, they are held back by a constraint that is difficult to overcome. This affliction plagues the most common form of microLED, which is made from InGaN and related materials, and relies on growth and processing conditions that demand rigid crystalline substrates, such as sapphire. While such substrates provide the materials quality that’s required for efficient emission, they fix the device in a form that cannot bend or conform. Note, though, that it’s not possible to turn to flexible polymeric substrates that offer mechanical compliance, because they are incompatible with the fabrication processes needed to produce high-quality emitters. So, the crux of the matter is a fundamental mismatch: the platform delivering performance fails to enable functionality.

To overcome this limitation requires a radical re-think, regarding how microLED systems are assembled. It’s an issue we’ve been exploring at the University of Waterloo and PulseForge, with success coming from a pioneering process that separates device fabrication and system integration.

Our process begins with the optimisation of the microLED on its native substrate. These emitters are then transferred to a flexible platform using a two-step process that preserves electrical and optical performance (see Figure 2). Crucially, this approach exposes the device’s backside, enabling control over light extraction beyond the constraints of the growth wafer.

Figure 2. Scanning electron microscope image of microLED arrays on a flexible substrate after 2-part transfer process.

A rigid foundation
When combining different materials systems in a single device, the challenge is not just a matter of bringing disparate entities together; integration forces a tradeoff.

One option is to keep materials separate, and connect them only through wire bonds, or by mounting side-by-side on a circuit board. The merit of this approach is that it preserves material quality, as each device is fabricated on its ideal substrate. But there is a downside – a limitation on how tightly the systems can interact.

At the other extreme is the integration of materials directly at the device level. However, while this enables functions and scaling outside the realm of what’s possible with conventional interconnects, this level of integration comes with strict constraints: the materials must be chemically, thermally, and mechanically compatible. Break these non-negotiables and defects emerge, performance degrades, and the advantages of integration vanish.

The trade-off is particularly severe for III-nitride microLEDs. Production of high-performance InGaN devices demands growth and processing under conditions that require rigid, crystalline substrates, such as sapphire. These substrates are great at providing the thermal stability and structural quality needed for efficient light emission. That’s not the case with flexible materials, which cannot support the temperatures or processes needed to build high-quality devices, but have the form required for conformal systems. So, the rub is that the platform that delivers performance is not the platform that enables flexibility.

Our approach, which separates these roles, is conceptually simple. However, it’s difficult to execute, because unlike packaged chips, microLEDs at the same stage are exposed and fragile. This issue arises because shrinking device dimensions to the micron scale leads to an increase in sensitivity to mechanical stress, thermal exposure, and process variation (see Figure 3). Note that techniques that are effective for larger devices, such as conventional pick-and-place assembly, do not readily scale to dense arrays of micronscale emitters.

Figure 3. Scanning electron microscope image of a 6 µm microLED before transfer.

So, the obstacles we have faced are not limited to figuring out how to simply move devices from one surface to another. While doing so, it’s imperative that we retain the performance that’s realised on the native wafer, while integrating devices onto a new platform that enables additional functionality. Our two-step process, fulfilling those objectives, is summarised in Figure 4.

Figure 4. MicroLEDs are first grown and fabricated on a rigid native substrate, such as sapphire, to support high-temperature III-nitride processing. Completed devices are then bonded to a temporary carrier, to allow the removal of the native growth substrate. After backside access and optional optical engineering, the microLEDs are bonded to a thin flexible polymeric substrate. A final photonic die transfer step separates the temporary carrier, leaving intact high-performance microLEDs on a mechanically flexible platform.

Photonic die transfer
A key part of our process involves separating where the device is made from where it is used. This step, introduced only at the final stage – it’s after the devices have been fully defined – ensures retention of established fabrication processes.

The release of microLEDs from their native substrate is undertaken with laser lift-off. This is a well-established technique in III-nitride processing, used to remove sapphire while preserving device integrity. During this step, we transfer devices onto a temporary carrier, which enables subsequent handling and processing.

Where the biggest challenge lies is in the detaching of devices from their growth wafer, and their subsequent transfer to a flexible substrate, a move that must be made without degrading performance. For micron-scale emitters arranged in dense arrays, this is far from straightforward. Flexible polymeric substrates are unable to withstand large mechanical forces, prolonged heating, or aggressive chemical processing. This rules out common microLED transfer techniques, such as laser, thermal, mechanical, and elastomer-based methods, as they involve stress, non-uniform separation, or excessive thermal exposure. These effects hamper performance by increasing leakage current, degrading contacts and reducing emission uniformity. Ultimately, there’s a lowering of the yield of functional devices.

We overcome these limitations with a form of photonic die transfer. Based on PulseForge technology, our approach allows the transfer of devices from the carrier onto the flexible substrate. Rather than relying on mechanical separation or bulk heating, we direct short broadband light pulses through a transparent carrier to an engineered bond interface (this is illustrated in Figure 5). Optical energy is converted into highly localised heat through an embedded absorbing layer, leading to separation precisely at the interface. Thanks to confinement of energy at the bond layer, we minimise mechanical shock and limit thermal exposure to the microLEDs.

Figure 5. PulseForge assisted photonic die transfer uses short broadband light pulses to expose the carrier light-absorbing layer, driving rapid heating and carrier release without direct mechanical stress to the microLEDs. The process supports large-area operation and allows the temporary carrier to be recovered for reuse.

Our localised, interface-driven release distinguishes our photonic die transfer technology from conventional transfer methods. We avoid mechanical and peel-based techniques, which apply stress across the entire device and increase the risk of fracture in micron-scale emitters; and we also steer clear of thermal release methods, which heat the entire stack – this is potentially damaging to temperature-sensitive materials, limiting compatibility with polymeric substrates, and increasing overall processing time. Instead, thanks to the technology developed at PulseForge, we selectively deliver energy only where separation is required, enabling rapid, non-contact transfer while preserving device integrity. Our process can be scaled across large areas, by stepping the light source (see Figure 6). This supports parallel transfer of dense microLED arrays, as we don’t have to handle devices one at a time.

Figure 6. The PulseForge transfer system enables large-area processing by stepping the light source across the substrate, supporting the parallel transfer of dense microLED arrays. The process supports large-area operation and allows the temporary carrier to be recovered for reuse.

An additional advantage with our approach is the opportunity to reuse the temporary carrier. As the separation occurs at the bond interface, and leaves the carrier largely unaffected, this can be recovered and re-used in subsequent cycles with minimal rework. Benefits are a trimming of materials consumption, a significant reduction to process cost, and an addressing of the sustainability challenge in advanced semiconductor manufacturing.

With conventional transfer approaches, these benefits are realised rarely. Often sacrificial layers or carrier substrates are discarded after a single use, contributing to materials waste and an increased processing overhead.

Design after growth
Our photonic die transfer process does more than move microLEDs onto a flexible substrate. It also removes a key limitation that is built into devices on their native wafer: a significant fraction of the generated light never escapes.

On sapphire, photons emitted in the active region are partly guided into the substrate, where they are lost, either through trapping or absorption. These optical losses are set by the growth geometry and cannot be addressed after fabrication. Removing sapphire eliminates this constraint and exposes the GaN backside, as shown in Figure 2, to create a new optical boundary that can be engineered before final integration.

This matters because microLED efficiency is not determined solely by photon generation. It also depends on how effectively photons exit the device. When the backside is accessible, the optical environment can be modified during the transfer sequence, rather than being fixed by the growth wafer.

The immediate benefit is a reduced substrate-guided loss, allowing more light to escape for the same electrical input. Ultimately, there’s an increase in effective brightness.

Further gains could be realised by introducing a reflective layer or textured surface on the exposed device backside prior to bonding to the flexible substrate. This layer would redirect photons otherwise propagating away from the emission surface, returning them toward the viewer. In this way, light previously lost into the substrate would be partially recovered, increasing the usable output without altering the diode structure itself.

This shift changes the role of transfer. On the native wafer, optical losses are largely fixed once growth is complete. But after substrate removal, they become adjustable. So, the same step that enables mechanical flexibility also creates an opportunity to improve light extraction, provided the device and its optical surroundings are considered together.

To determine whether this added flexibility comes at the expense of device integrity, we have compared performance before and after integration. We initially characterised microLEDs on sapphire, to establish baseline electrical and optical behaviour, prior to transfer and release onto the flexible platform using photonic die transfer. Any degradation would appear as increased leakage current, a shift in emission spectra, or a reduced efficiency. But we find that devices retain their electrical and spectral characteristics while benefiting from the modified optical environment, confirming that transfer preserves device quality while enhancing light extraction.

Preserved quality
By separating fabrication from integration, we have introduced a potential pitfall: Could the transfer process degrade the device?

MicroLEDs are sensitive to defects, contact damage, and thermal and mechanical stress – and all have an immediate impact on the electrical response. If there’s an increase in leakage current, this indicates defect formation in the junction, while changes in forward bias behaviour point to contact degradation or increased series resistance. There might also be a small uptick in turn-on voltage, raising power consumption and reducing operating efficiency.

Due to these concerns, we measure our device performance before and after transfer. By scrutinising current-voltage characteristics, emission spectra, and efficiency, we obtain a direct comparison of device integrity through the process (see Figure 7).

Figure 7. (a) Current voltage characteristics for 6 µm microLEDs show comparable behaviour on sapphire and after transfer to flexible substrates, including under flat, tensile, and compressive bending conditions. (b) Normalised electroluminescence spectra remain stable under bending, with only minor wavelength shifts and no measurable spectral broadening.

With our approach, the electrical response does not change. Our microLEDs-on-sapphire exhibit a low leakage current and stable forward operation, consistent with high-quality junctions. Transfer to a flexible substrate preserves these characteristics, even for devices as small as 6 µm in diameter. Due to an absence of increased leakage, alongside no alterations to forward behaviour, we can deduce that our photonic die transfer steps do not introduce measurable electrical damage.

What’s equally encouraging is that our optical response shows the same stability. Emission spectra remain unchanged after transfer, with no measurable shift in peak wavelength (see Figure 7 (b)). So we can conclude that the active region is unaffected by transfer, with recombination processes governing light generation remaining intact.

Another matter must be investigated with the introduction of mechanical flexibility: Does performance remain stable under deformation? We have found that under tensile and compressive bending, the emission profile shows negligible change. This stability is essential. Even small spectral shifts would limit the usefulness of microLEDs in biomedical systems (and also in displays and optical communication), due to the need for a consistent colour and wavelength during operation.

Based on our thorough examination, we conclude that our transfer process introduces no measurable electrical or optical penalty. Our microLEDs retain their behaviour on the native wafer after substrate removal, integration, and mechanical deformation. By preserving performance, if there is any change to brightness or efficiency, we attribute this to a change in the optical environment, rather than damage to the device.

Unlocked efficiency
A really important question to address is whether the performance of our microLEDs improves after they are removed from the growth substrate.

Measurements reveal that efficiency increases. For 6 µm microLEDs, peak wall-plug efficiency rises from around 9 percent on sapphire to approximately 12 percent on the flexible substrate (see Figure 8). We have also studied devices with other dimensions, finding that peak efficiency improves by up to 1.5 times. This boost in efficiency is beneficial, as it reduces the power required to realise the same brightness.

It should be emphasised that this gain in efficiency is not driven by the diode. The electrical characteristics are unchanged, with emission spectra showing no measurable wavelength shift. Carrier injection and radiative recombination are also unaffected. What’s behind the improvement is how the light is extracted.

Figure 8. Wall-plug efficiency as a function of absolute current density shows higher peak efficiency after transfer to the flexible substrate, consistent with increased light extraction.

The removal of sapphire eliminates a major loss pathway, allowing fewer photons to be trapped. In conjunction, exposing the GaN backside changes optical boundary conditions set by the growth wafer. For our microLEDs, this change alone increases the fraction of usable light. In future, we could investigate more advanced designs that build on this, using backside texturing or reflective layers to extract more photons that would otherwise be lost.

Separation is key. While growth defines how efficiently photons are generated, transfer determines how effectively they are delivered.

The gains in efficiency that come from separation are more pronounced as device dimensions shrink. For microLEDs below 10 µm, efficiency tends to decline, due to an increase in surface recombination. But if that’s avoided, and efficiency is maintained while scaling, this enables higher pixel densities, finer spatial control, and lower power operation. They are significant benefits, as these characteristics are essential for wearable and implantable biomedical systems, where optical sources must deliver precise output on soft, moving surfaces without excessive heat.

Beyond bending
Once the microLEDs are free from rigid substrates, they can be combined with platforms that match the mechanical and functional demands of the application. This includes new opportunities to control light extraction. As well as flexibility, there’s the opportunity to use light more effectively. That’s a particularly valuable combination in systems demanding precise optical delivery within a limited form factor, such as wearable and implantable biomedical platforms that must operate reliably on soft, moving surfaces.

While our success has focused on microLEDs, our approach is applicable to other forms of compound semiconductor devices that face the same constraint – that is, the substrate that enables optimal growth is not the one that best supports system-level operation. Transfer-based integration provides a way to resolve this mismatch, allowing devices to be fabricated under ideal conditions, before placing them where they deliver the greatest impact.

This changes the role of transfer. It is no longer just a final packaging step, but part of device design. It’s now possible to engineer electrical performance, optical behaviour, and mechanical function separately, prior to integration.

For biomedical systems, this enables a new class of compact, high-resolution optical platforms, where dense arrays of microLEDs deliver precise, stable light on conformal substrates with reduced power consumption. By building devices on wafers optimised for performance, prior to their assembly into systems that maintain that performance while operating directly on the body, our technology is bringing continuous, high-fidelity optical interaction into practical wearable and implantable systems.