1. INTRODUCTION
Micro-LED technology is widely considered one of the most promising candidates for
next-generation display systems[1-
4]. Compared with conventional liquid crystal display (LCD) and organic light-emitting
diode (OLED) technologies, micro-LEDs offer several key advantages, including high
brightness, improved energy efficiency, and excellent thermal and operational stability[5]. As micro-LEDs are based on inorganic semiconductor materials rather than organic
emissive layers, they can achieve higher luminance and longer device lifetimes. In
addition, each micro-LED chip functions as an individual pixel, enabling extremely
high-resolution displays with superior reliability[2].
Owing to these advantages, micro-LED technology has attracted increasing interest
for a wide range of applications, such as large-area televisions, wearable devices,
automotive displays, and emerging augmented reality systems[6]. Despite these advantages, several technical challenges remain before micro-LED displays
can be fully commercialized[7].
One of the most critical challenges is the mass transfer of the microscale LED chips
from the growth wafer to the target substrate[8]. In typical micro-LED fabrication processes, individual LED chips are first fabricated
on an epitaxial wafer and then transferred to a display backplane[5].
For large-area displays, this process requires the accurate placement of millions
of microdevices. Consequently, the transfer process must achieve both high alignment
precision and high throughput, while maintaining excellent process reproducibility[9].
Various microdevice transfer approaches have been explored to address this issue,
including elastomer stamp transfer, electrostatic transfer, and fluidic self-assembly.
Among these approaches, laser-induced forward transfer (LIFT) has emerged as a promising
technique owing to its noncontact nature and high spatial precision[10]. In the LIFT process, pulsed laser irradiation generates localized energy at the
interface of the donor substrate, enabling the forward transfer of materials or devices
toward the receiver substrate[11,
12]. As the transfer process can be controlled by adjusting the laser parameters, LIFT
has been investigated as a potential method for microdevice assembly and micro-LED
transfer applications[13].
Laser-based transfer techniques have been extensively studied for microscale device
assembly and material processing[14]. Early LIFT studies mainly focused on transferring functional materials, such as
metals, polymers, and inks, to form micropatterns or electronic structures[15,
16]. As the technology progressed, LIFT began to be applied to the transfer of microdevices
and semiconductor components. More recently, several studies have explored the use
of LIFT for micro-LED device transfer, demonstrating that laser-driven processes can
enable the precise positioning of microscale devices and offer a potential route for
large-area device integration[17,
18].
However, many previous studies have focused on demonstrating the feasibility of device
transfer under specific experimental conditions[10]. Systematic investigations of the process parameters and their influence on transfer
behavior remain relatively limited[19]. In addition, most studies have been conducted using a single device size or within
a narrow process window, making it difficult to establish reliable process references
for practical applications. From a manufacturing perspective, micro-LED display production
requires a transfer process that can achieve both high yield and practical throughput[19]. As the display resolution increases, the number of microdevices required for a single
panel significantly increases. Therefore, even a small reduction in transfer yield
can result in substantial production losses. Establishing a robust and scalable transfer
process is essential to bridge the gap between laboratory-scale demonstrations and
practical micro-LED manufacturing technologies.
In this study, we investigate a systematic approach for LIFT-based microchip transfer
by integrating temporal and spatial energy considerations. This study analyzes the
relationship between laser frequency, temporal irradiation condition, and effective
energy delivery to examine the effect of frequency on transfer behavior.
In addition, the geometric relationship between the laser spot size and chip size
is introduced as a key parameter governing the spatial energy distribution, enabling
stable transfer behavior under the tested chip-size conditions. Based on these considerations,
optimized process conditions are experimentally identified and validated under the
present experimental conditions for microchips with sizes ranging from 30 to 70 µm,
achieving transfer yields exceeding 99% with high reproducibility. These results provide
useful insights into the relationship between process parameters and transfer behavior.
The findings suggest that appropriate control of temporal and spatial parameters can
serve as a useful practical reference for optimizing LIFT-based microchip transfer
under similar experimental conditions.
2. EXPERIMENTAL
Silicon microchips with lateral sizes ranging from 30 to 70 µm were prepared on a
glass carrier coated with a polyimide-based dynamic release layer (DRL). The DRL thickness
was controlled to approximately 1 µm based on previously reported blister-driven transfer
conditions under similar UV laser irradiation environments[12].
A PDMS substrate was used as the receiver, and no intentional gap was introduced between
the donor and receiver substrates during the transfer process. Blister-driven transfer
was performed using a pulsed ultraviolet (UV) laser with a wavelength of 355 nm.
The laser frequency was varied in the range of 100–260 kHz, while the pulse duration
was fixed at 3.8 μs based on the laser system setting. The pulse energy was maintained
at approximately 23.5 µJ under the present experimental conditions, while the effective
energy density varied depending on the spot size condition[12]. Although the main frequency analysis was conducted within the range of 100–260 kHz,
an additional low-frequency condition at 50 kHz was examined for the 70-µm chips to
investigate the effect of increased temporal energy delivery under larger chip conditions.
The frequency and spot size were systematically varied to investigate their effects
on the transfer stability. The spot-to-chip size ratio was adjusted by controlling
the irradiation condition, resulting in different spot sizes relative to the chip
size.
Laser scanning was performed under single-exposure conditions at a fixed scanning
speed of 500 mm/s. During the experiments, only one parameter was varied at a time,
while all other process parameters were kept constant.
The transfer yield was evaluated based on positional alignment and structural integrity
after transfer. Reproducibility was assessed through five repeated trials under identical
process conditions.
3. RESULTS AND DISCUSSION
3.1. Operational Frequency Window
Laser frequency is a key parameter governing the stability of LIFT processes[20]. In scanning-based LIFT systems, the transfer behavior is strongly governed by the
temporal delivery of the laser energy to the chip. Among the various parameters, the
laser frequency directly determines the number of pulses interacting with the chip
during the scanning process, making it a critical factor for controlling energy accumulation
and transfer stability[21].
In scanning-based LIFT systems, the effective interaction between the laser beam and
an individual chip is determined not only by the laser energy but also by the number
of laser pulses interacting with the chip during the scanning process[19].
Therefore, the frequency directly controls the temporal energy delivery and plays
a critical role in determining transfer stability[21]. Under the scanning conditions, the effective interaction time between the laser
beam and a single chip can be approximated as
The interaction time determines the duration for which a single chip is exposed to
the laser beam during scanning, which directly affects the total energy delivered
to the chip[20]. In Eq. (1), L is the chip size and v is the scanning speed. The scanning speed was fixed at
500 mm/s, and the chip size used in this analysis was 50-μm. This approximation is
widely used to describe the laser–material interaction time in scanning laser systems.
For the present experimental conditions, the chip size was $L = 50$ µm and the scanning
speed was $v = 500$ mm/s, resulting in an interaction time of
The interaction time determines the number of laser pulses that interact with a single
chip during scanning[10].
The number of laser pulses interacting with a chip can therefore be expressed as
This indicates that increasing the frequency increases the number of pulses contributing
to the energy delivery, thereby potentially increasing the total energy supplied to
the chip[10]. In Eq. (3), $f$ represents the laser frequency. Increasing the frequency increases the number
of pulses delivered to the chip. To describe the temporal irradiation condition, a
dimensionless parameter is defined as
where τ is the effective laser irradiation duration. In this study, τ was set to 3.8
μs based on the laser system setting. As the frequency increases, the fraction of
laser-on time within a given period also increases, which may lead to enhanced energy
accumulation within the dynamic release layer. This accumulated energy can influence
the thermal response of the release layer and affect the transfer behavior. Based
on these considerations, a simplified phenomenological expression was introduced to
qualitatively describe the competing effects between effective pulse delivery and
excessive temporal overlap during the transfer process:
In this expression, the term (1 − δ(f)) represents the reduction in transfer stability
caused by excessive temporal irradiation and thermal accumulation at high frequencies.
Therefore, transfer stability can be influenced by the balance between sufficient
pulse delivery and excessive temporal irradiation or energy accumulation, which can
lead to thermal damage. Substituting the expressions for $N(f)$ and $\delta(f)$ defined
in Eqs. (1)–(3), the expression can be rewritten as
Because $L/v = 10^{-4}$ under the present experimental conditions, Eq. (6) can be simplified as
To estimate a characteristic frequency, the derivative of the expression is calculated
as
Setting the derivative to zero yields a characteristic frequency
For the pulse duration used in this study ($\tau = 3.8$ µs), a characteristic frequency
can be estimated as
To compare with this simplified analysis, transfer experiments were conducted at frequencies
of 100, 150, and 260 kHz under a fixed scanning speed of 500 mm/s with no intentional
gap between the donor and receiver substrates. A pulsed UV laser (λ = 355 nm) with
a pulse duration of 3.8 μs was used, and all other process parameters were kept constant
during the experiments. Under these fixed conditions, at 100 kHz, the δ value was
relatively small ($\delta = 0.38$), resulting in insufficient pulse delivery to sustain
stable transfer.
This resulted in unstable transfer behavior and reduced yield. At 260 kHz, the δ value
approached unity ($\delta \approx 0.99$), indicating excessive temporal irradiation.
In this regime, energy accumulation becomes dominant, reducing the transfer stability
and transfer yield. By contrast, the highest transfer yield and most stable transfer
behavior were observed at 150 kHz.
This frequency is close to the characteristic value estimated from the simplified
analysis (f* ≈ 132 kHz), indicating that the experimentally observed optimum at 150
kHz is generally consistent with the trend suggested by the simplified analysis. The
discrepancy between the estimated characteristic frequency and the experimentally
observed optimum is attributed to the simplified nature of the phenomenological model.
The model does not account for factors such as spatial beam nonuniformity, thermal
diffusion within the DRL, pulse-to-pulse thermal accumulation, and material-dependent
absorption behavior. Although only three representative frequencies were examined
in this study, the results suggest that an intermediate frequency condition provides
more stable transfer under the present experimental conditions. This suggests a practical
guideline for selecting appropriate frequency conditions under similar experimental
setups[22].
Furthermore, the results highlight that frequency is intrinsically coupled with temporal
energy accumulation and that a narrow process window exists for stable transfer. From
a process perspective, these results suggest that transfer behavior is influenced
by the balance between pulse number and temporal irradiation condition. This provides
a useful practical reference for selecting appropriate frequency conditions under
similar experimental setups.
Fig. 1. (a) Schematic of the LIFT process. (b) 2D optical image of the 50-µm chip
LIFT experiment. (c) Number of transferred chips as a function of laser frequency.
(d) Number of successfully transferred chips excluding defects as a function of laser
frequency. (e) Transfer yield as a function of laser frequency. (f–h) Optical microscopy
images of the transfer results at 100, 150, and 260 kHz.
3.2. Spot-to-Chip Size Ratio
For the 70-µm chips, this larger chip size naturally requires a higher amount of energy
to initiate transfer compared to the 50-µm case. From a physical perspective, the
total delivered energy and spatial distribution of energy across the chip play a critical
role in determining the transfer stability[22]. Therefore, the spot size relative to the chip size is an important parameter for
controlling the energy distribution. Based on the relationship derived in Section
3.1, the initial experiments were performed at 50 kHz under the focused condition
(i.e., zero defocus distance, where the laser focal plane coincides with the PI layer
surface). However, under this condition, most of the chips were severely burned and
stable transfer was not achieved. This result suggests that increasing the delivered
energy alone does not lead to a stable transfer[23]. Instead, the manner in which the energy is distributed over the chip area becomes
increasingly critical, especially as the chip size increases.
To address this issue, the spot size was increased to reduce the effective energy
density applied to the chip. To describe this effect, the spot-to-chip size ratio
was defined as
where $D_{spot}$ is the laser spot diameter and $D_{chip}$ is the chip size. This
ratio provides a normalized parameter that describes how the laser energy is spatially
matched to the chip geometry under the present experimental conditions. When the spot
size was approximately 43 µm, some of the chips were transferred, but most still showed
thermal damage similar to the focused condition. The transfer behavior in this regime
was not uniform, indicating that the energy distribution across the chip remained
uneven. This indicates that a locally concentrated energy distribution can lead to
partial damage even when the total delivered energy is sufficient for transfer. As
the spot size was increased to approximately 51 µm, the transfer behavior noticeably
changed. Most of the chips were transferred without visible damage, and the transfer
yield exceeded 95%[9]. Under these conditions, the energy appeared to be distributed more evenly across
the chip, avoiding a local energy concentration. The corresponding ratio was
This result suggests that stable transfer was achieved under the present 70-µm chip
conditions when the spot size became comparable to the chip size, ensuring a more
uniform spatial energy distribution. When the spot is too small, the energy is concentrated
in a limited region, leading to damage[9]. However, when the spot size is increased, the energy spreads over a wider area,
allowing for a more uniform interaction with the chip. Therefore, the spot-to-chip
ratio can be considered as a useful parameter that governs the balance between energy
concentration and distribution across the chip surface. Thus, the spot-to-chip ratio
provides a simple method for describing how well the laser energy matches the chip
geometry.
These results suggest that a ratio of approximately 0.7 was found to provide stable
transfer under the present experimental conditions[22]. It should be noted that the α ≈ 0.73 condition was derived specifically from the
70-µm chip experiments under the present process conditions. Therefore, this value
should be regarded as a practical reference rather than a universal criterion. It
should be noted that this observation is based on the tested chip size (70 μm) and
experimental conditions used in this study. From a process perspective, this parameter
is useful because it can be used as a practical reference for similar chip sizes under
comparable experimental conditions.
Fig. 2. (a–c) Optical microscopy images for spot sizes of 40, 43, and 51 µm. (d) Optical
microscopy image used for single spot size measurements. (e) Schematic of the spot-size-to-chip-size
experiment. (f) Percentage ratio of the spot size to the chip size. (g) Number of
transferred chips as a function of the spot size. (h) Transfer yield as a function
of the spot size.
3.3. Optimization Results for Different Chip Sizes
Based on the optimized conditions identified from the frequency and geometric considerations
discussed in the previous sections, transfer experiments were carried out for chip
sizes of 30, 50, and 70-µm. This experiment was designed to verify whether the optimized
conditions could be applied under the tested chip-size conditions, which is important
for practical applications. Overall, stable transfer behavior was observed across
all chip sizes. The transfer yield exceeded 99% in all cases[24]. Even for the larger 70-µm chips, a reasonably high yield was maintained once the
spot size was properly adjusted[25,
26].
This further confirms that spatial energy distribution plays a dominant role in maintaining
the transfer stability of larger chips. It should be noted that the 30-µm chips showed
stable transfer behavior under conditions similar to those used for the 50-µm chips.
For the 30- and 50-µm chips, stable transfer was also achieved under comparable spatial
energy distribution conditions, although the corresponding α values were not independently
optimized.
This suggests that, under the present experimental conditions, the transfer process
becomes less sensitive to the absolute chip size once the energy distribution condition
is satisfied. This can be attributed to the spatial energy distribution of the laser
beams. Because the spot size is larger than the chip size, sufficient energy is delivered
to the entire chip area, whereas the Gaussian profile prevents excessive energy concentration
in specific regions.
Another factor that contributes to stable transfer is the spacing between neighboring
chips. Because the chips were sufficiently separated, thermal or mechanical interactions
between adjacent chips were minimized. This indicates that interchip interference
is negligible under the current process conditions, allowing for independent and uniform
transfer behavior across the array. As a result, each chip behaves independently during
the transfer process, which helps to maintain uniformity across the array.
Collectively, these results suggest that the optimized conditions were applicable
to the tested chip sizes. The transfer behavior showed a similar trend across the
tested chip sizes, governed by the balance between energy delivery and its spatial
distribution.
It should be noted that these observations are limited to the tested chip sizes (30–70
μm) and experimental conditions used in this study.
In practical terms, these results indicate that similar transfer behavior can be achieved
for the tested chip sizes by appropriately adjusting the process conditions. This
suggests that the relationship between frequency and spot-to-chip size ratio can be
used as a useful reference for process optimization under comparable experimental
conditions.
Fig. 3. (a–c) Optical microscopy images of the optimized transfer results for chip
sizes of 30, 50, and 70 µm, respectively. (d) Schematic of the chip-size-dependent
transfer experiment. (e) Number of transferred chips as a function of chip size. (f)
Transfer yield as a function of chip size.
3.4. Process Reproducibility Evaluation
Process reproducibility was evaluated to determine whether the optimized LIFT conditions
could maintain stable transfer behavior under repeated operations. Reproducibility
is an important requirement for reliable process operation because consistent performance
under identical conditions directly determines process reliability[27,
28]. In practical applications, achieving a high yield in a single experiment is insufficient,
and the process must consistently reproduce the same performance under identical conditions[29]. Transfer experiments were repeated five times under the optimized conditions using
50-µm chips. Across all repetitions, the transfer behavior remained highly consistent
with no noticeable variation in chip alignment or structural integrity. This indicates
that the process is robust against minor fluctuations in the experimental conditions.
Moreover, the transferred chips exhibited uniform positioning and stable transfer
characteristics, further confirming the stability of the transfer behavior under the
same conditions[8]. This low variation demonstrates that the optimized process operates within a stable
process window. Similarly, the number of successfully transferred chips showed a negligible
deviation across repetitions. These results demonstrate that the optimized process
provides highly stable and repeatable performance. Figure 4(c) schematically summarizes the relative influence of the investigated process parameters
based on the experimentally observed transfer trends. The ranking is intended as a
qualitative interpretation rather than a quantitative statistical analysis. Analysis
of the process parameter contributions further revealed that frequency most significantly
influenced the transfer stability, followed by the spot-to-chip size ratio and chip
size. This hierarchy indicates that temporal energy control plays a dominant role,
whereas spatial and geometric factors provide additional stabilization. In particular,
the chip size influences the energy distribution required for stable transfer. Larger
chips require a broader and more uniform energy profile, which makes them more sensitive
to spatial energy conditions. This observation is consistent with the results presented
in Section 3.3, where stable transfer across different chip sizes was achieved only
when the spot size was appropriately scaled relative to the chip size.
From a physical perspective, the high reproducibility can be attributed to the fact
that both the temporal and spatial energy conditions are maintained within a stable
operating window. Because the balance between pulse delivery and energy distribution
is preserved, the transfer behavior remains consistent across repeated trials.
These results demonstrate that the optimized conditions enable stable and repeatable
transfer behavior. More importantly, the results confirm that the transfer process
is governed by the coupled effects of the temporal parameters, spatial energy distribution,
and geometric scaling. These observations provide useful insights for understanding
the transfer behavior under the present experimental conditions. From a practical
perspective, this reproducibility is important for achieving consistent transfer behavior
under repeated operations[30,
31].
Fig. 4. (a) Schematic of the reproducibility experiment. (b) Optical microscopy image
of the reproducibility experiment results. (c) Relative importance of the process
parameters considering their effect on the transfer yield. (d) Transfer yield as a
function of repeated experiments. (e) Number of transferred chips as a function of
repeated experiments.