Though necessary and advantageous in many fields, the high coherence of lasers is detrimental to their performance in certain applications, including illumination, imaging, and projection. This is due to the formation of coherence artifacts, commonly known as speckles, resulting from the interference of randomly scattering spatially coherent photons. It is possible to resolve this issue by increasing the number of mutually incoherent modes emitted from the laser. In vertical-cavity surface-emitting lasers (VCSELs), this can be performed by designing them to have chaotic cavities. This paves the way toward their use in simultaneous illumination and communication scenarios. Herein, we show that chaotic-cavity broad-area VCSELs can achieve significantly broader modulation bandwidths (up to 5 GHz) and higher data rates (up to 12.6 GB/s) compared to other low-coherence light sources, with a lower speckle contrast. We further report a novel technique for lowering the speckle contrast by carefully designing the AC signal used for communication. We show that the apparent spatial coherence is dramatically decreased by inserting a short chirp signal between symbols. Using this method with a chaotic-cavity VCSEL, the number of apparent modes can be up to 450, compared to 88 modes measured from a conventional broad-area VCSEL (a fivefold increase). In light of the recent advances in visible-light VCSELs, this work shows the potential of low-coherence surface-emitting lasers (LCSELs) in simultaneous illumination and optical wireless communication systems since they combine the high speed of lasers with the excellent illumination properties of light-emitting diodes.

Spatial and temporal coherence of light are needed in a wide range of applications, including fiber-optic communication, spectroscopy, laser cooling, and atomic clocks. However, in other applications, the high degree of coherence degrades the performance of the laser-based system due to the inevitable formation of unintended interference artifacts.1 For example, in laser-based imaging, illumination, and display and projection systems, the random scattering of spatially coherent photons through the propagation medium results in random constructive and destructive interference, which results in the formation of speckles.2,3 Moreover, the long coherence length (high temporal coherence) of lasers limits the resolution of time-domain optical coherence tomography as well as other interferometric sensing systems.4 For these applications, a lower degree of coherence is always desirable.

Due to their advantages in the aforementioned applications, different types of low-coherence light sources have been developed over the years. Broadband edge-emitting lasers based on quantum dot/dash active regions have been demonstrated5–7 (although single-wavelength emission is possible using tunnel-injection structures8). Superluminescent diodes (SLDs), which rely on suppressed optical feedback to prevent the formation of longitudinal modes, have also been shown to exhibit low temporal coherence, characterized by their wide, smooth emission spectra.9–13 However, SLDs operate with high spatial coherence because of the narrow-waveguide design they use.1,13 It is also possible to embed colloidal quantum dots in random photonic crystal cavities to achieve emission with lower coherence.14 Furthermore, chaotic-cavity and random edge-emitting lasers have been developed as sources of low-spatial-coherence light.15,16 Chaotic-cavity lasers rely on the chaotic ray dynamics within these cavities to prevent the formation of a few modes with high quality factors.15 This allows more modes to lase simultaneously in the cavity, lowering the spatial coherence and enabling speckle-free illumination. Nevertheless, the above-mentioned devices use horizontal waveguides and edge-emitting configurations. Therefore, they lack the advantages offered by surface emission from light-emitting diodes (LEDs) and vertical-cavity surface-emitting lasers (VCSELs).

Since their invention in the 1970s,17–19 VCSELs have seen a surge in interest after their implementation in consumer electronics, especially for light detection and ranging (LiDAR).20 Their low manufacturing cost, two-dimensional array configuration, and high beam quality made them an ideal source for many applications and led to their widespread use. In addition, with the recent advancements in the development of visible-light VCSELs,21–23 they have the potential to be used in more applications. Especially of interest here is using them in simultaneous illumination and optical wireless communication (OWC) systems, also known as light-fidelity (Li-Fi).

In Li-Fi, wireless signals can be transmitted through light used for indoor illumination, making these systems highly efficient.24 Moreover, they provide additional security features since the light in the optical spectral range cannot penetrate walls. Because of these advantages, Li-Fi has received significant research interest in recent years.25–27 However, one of the main limitations of Li-Fi is the strict alignment requirements needed to ensure stable, high-speed communication. This can be circumvented using arrays of transmitters, which are easily achievable using LEDs and VCSELs. However, LEDs are less directional and cannot achieve the high speed of lasers, which makes the use of VCSELs more suitable. In this context, a VCSEL array can be used with space-division multiplexing (SDM) and angle diversity in indoor attocell networks,28,29 in which each VCSEL is assigned a small cell in a room and acts as its base station. The signal follows the mobile device as it moves by hopping from one VCSEL to another, achieving the functionalities of beamforming and beam steering without the need for optical phased arrays. Furthermore, near-infrared (NIR) VCSELs can also be used for the up-link in Li-Fi systems. However, conventional VCSELs suffer from the disadvantages caused by their relatively high spatial coherence, which results in speckles and poor illumination quality. This is especially the case when each individual VCSEL is illuminating a specific area.

Recently, we showed that by using chaotic-cavity broad-area VCSELs, the spatial coherence can be substantially reduced.30 This comes with an increase in optical power of up to 66.7%. This is because of the lower quality factor of the modes supported by these cavities. The performance of small D-shaped VCSELs in fiber-based communication has been previously tested.31 However, the small aperture size and the short distance between the flat edge and the center of those reported devices reduce the chaotic behavior in the cavity.15,30 Herein, the potential of chaotic-cavity broad-area VCSELs in OWC is explored for the first time. We also demonstrate a novel technique to lower the apparent coherence of the emitted light from these VCSELs by tailoring the AC signal. Adding a short chirp signal in between communication symbols is shown to effectively lower the spatial coherence by increasing the number of modes of chaotic-cavity VCSELs by more than threefold (up to 454 modes), resulting in a speckle contrast of around 4.7%. The number of modes is more than five times higher than that of conventional VCSELs fabricated on the same sample. Moreover, the achieved speckle contrast is much lower than that reported using SLDs (20%).32 Finally, the modulation bandwidth and the achieved data rate (5 GHz, 12.6 GB/s) are significantly higher than the record values achieved using SLDs (2.5 GHz, 4.6 GB/s)33,34 while maintaining a higher optical power. Therefore, despite the fact that the lower coherence can come at the cost of having a higher relative intensity noise (RIN), degrading the communication performance, we demonstrated that chaotic-cavity VCSELs can still achieve relatively high-speed communication with improved illumination performance.

Two different types of VCSELs were fabricated on the same 940-nm VCSEL wafer. The epitaxial layer structure is shown in Figs. 1(a) and 1(b), which show a bright-field transmission electron microscope (TEM) image of the layers around the active region. The bright region is the oxidized part, and the darker stack in the center is the active region, as can be seen from the energy-dispersive X-ray spectroscopy (EDX) plots in Figs. 1(c)1(f), which show the atomic fractions of O, P, In, Al, Ga, and As and were obtained by averaging horizontal pixels close to the right edge of the TEM image in (b). The top gray stripe across the plots is aligned with the oxide aperture layer. The atomic fraction of O increases dramatically while that of As drops, verifying the oxidation of the intended layer. The second gray stripe marks the active region. A clear increase in the atomic fractions of In and P can be seen, while that of Al drops.

FIG. 1.

(a) The epitaxial layer structure of the VCSEL wafer used to fabricate the conventional and chaotic-cavity VCSELs. (b) A bright-field transmission electron microscope (TEM) image of the layers around the multiquantum well active region. (c)–(f) The atomic fraction of O, P, In, Al, Ga, and As at different depths obtained using energy-dispersive X-ray spectroscopy (EDX) aligned with the TEM image in (b). The top shaded region across the EDX plots (red arrow) marks the position of the oxide aperture, whereas the second shaded region (blue arrow) marks that of the active region. (g) and (h) show the light-output–current plots of the two conventional and the chaotic cavity VCSELs, respectively.39 The insets show the corresponding near-field profiles taken at the current values marked by the gray dots and the micrographs of the fabricated devices. (i)–(n) Illustrations of the fabrication process. The shown shape is for the conventional circular VCSEL.

FIG. 1.

(a) The epitaxial layer structure of the VCSEL wafer used to fabricate the conventional and chaotic-cavity VCSELs. (b) A bright-field transmission electron microscope (TEM) image of the layers around the multiquantum well active region. (c)–(f) The atomic fraction of O, P, In, Al, Ga, and As at different depths obtained using energy-dispersive X-ray spectroscopy (EDX) aligned with the TEM image in (b). The top shaded region across the EDX plots (red arrow) marks the position of the oxide aperture, whereas the second shaded region (blue arrow) marks that of the active region. (g) and (h) show the light-output–current plots of the two conventional and the chaotic cavity VCSELs, respectively.39 The insets show the corresponding near-field profiles taken at the current values marked by the gray dots and the micrographs of the fabricated devices. (i)–(n) Illustrations of the fabrication process. The shown shape is for the conventional circular VCSEL.

Close modal

Figures 1(i)1(n) show simplified illustrations of the steps of the fabrication process that was used. It includes the metallization of the p-type side, etching the mesas down to a depth of around 3.7 μm, defining the oxide aperture using wet oxidation (oxidation depth around 20–25 μm), depositing a silicon-nitride layer used for sidewall isolation, and depositing the metal pads for both the n- and p-type sides. More details of the layer structure and the fabrication process can be found in our previous work.30 

The two types of VCSELs have two different mesa shapes. The first is a conventional circular broad-area VCSEL (referred to here as an O-shaped or conventional VCSEL), whose aperture radius after oxidation is estimated to be around 28 μm. The second is a chaotic-cavity VCSEL designed to have a D-shaped cavity. It is defined by a circle with a flat cut made at a certain distance, d, from its center, where d can have any value from zero to the length of the radius of the circle, R. D-shaped cavities have been shown to result in chaotic ray dynamics in lasers.1,15 They have been used to reduce the spatial coherence of edge-emitting lasers and VCSELs.15,30 This was shown to be particularly effective when d = 0.5R.15 In our case, after the aperture oxidation, d is estimated to be around 0.42R. The two mesa shapes are designed such that the resulting aperture areas after oxidation are similar. The D-shaped VCSEL is estimated to have an area around 92% that of the circular VCSEL. Combined with the broad area of the VCSEL, the D-shaped cavity can support a large number of modes without resorting to designing even larger VCSELs, which would limit their speed.

The light-output–current (LI) plots of the two VCSELs are shown in Figs. 1(g) and 1(h) under continuous wave (CW) operation. The D-shaped VCSEL can achieve a higher maximum optical power (up to 36 mW). This is expected to be due to the more uniform optical field distribution in the active region, as shown from the near-field profiles of the VCSELs shown in the insets of Figs. 1(g) and 1(h). Broad-area circular VCSELs are known to support high-order modes, which are concentrated around the edge of the device, increasing the scattering losses and lowering the optical power due to unwanted reflections from the top metal contacts due to the beam divergence from the aperture. The increase in the optical power as well as the external quantum efficiency (EQE) have been previously demonstrated and studied in asymmetric-cavity VCSELs.30,35

We then measured the optical spectra of the two different types of VCSEL. The output light is collected using an integrating sphere connected to an optical spectrum analyzer (OSA) through a multi-mode, low-hydroxyl (low-OH) optical fiber whose core diameter is 200 μm and numerical aperture (NA) is 0.39. A low-OH fiber is chosen to avoid the non-uniform transmission of high-OH fibers in the range of 900–1000 nm, which covers the spectral range of the fabricated VCSELs. The spectra are shown in Figs. 2(a) and 2(b). The spectra were recorded using the maximum resolution of the OSA (20 pm) at different injection currents above threshold. The chaotic-cavity VCSEL has more components in its spectra due to the expected increase in the number of modes. This is because different transverse modes lase at slightly different wavelengths. This can be seen in the root mean square (RMS) spectral width extracted from these plots, which is shown in Figs. 2(c) and 2(d) for the conventional and chaotic-cavity VCSELs, respectively. The figures also show the average wavelength values, which exhibit a redshift with the increase in current due to Joule heating, which results in thermal expansion and hence a decrease in the bandgap of the active region. For the RMS spectral widths and average wavelength values to be more accurate, the noise below 8% of the peak of each spectrum (marked by the dashed gray line) is discarded.

FIG. 2.

(a) and (b) respectively show the optical spectra of the conventional and the chaotic-cavity VCSELs at different injection currents above threshold, while (c) and (d) respectively show their extracted average wavelengths and the root mean square spectral widths. Values below the dashed gray lines in (a) and (b) are considered noise for the calculations of the plots in (c) and (d). The insets show the shape of the device.

FIG. 2.

(a) and (b) respectively show the optical spectra of the conventional and the chaotic-cavity VCSELs at different injection currents above threshold, while (c) and (d) respectively show their extracted average wavelengths and the root mean square spectral widths. Values below the dashed gray lines in (a) and (b) are considered noise for the calculations of the plots in (c) and (d). The insets show the shape of the device.

Close modal

Although the RMS spectral width is a standard measurement for multi-mode VCSELs, it gives weight to the separation between different spectral components. Alternatively, calculating the area under the curve of each spectrum after being normalized by its maximum value gives a better indication of the increase in the number of modes, regardless of the differences between their wavelengths. To find the area under the curve at the maximum optical power, we measured the spectra of 6 pairs of VCSELs (12 VCSELs) at currents below threshold to above threshold with steps of 2 mA. The measurement was performed using a similar setup but by replacing the OSA with a faster spectrometer to avoid heating caused by the long scanning time of the OSA given the need to record tens of spectra for each VCSEL. These measurements and calculations verified the increase in the spectral components, as can be seen in Fig. S1 in the supplementary material. Nevertheless, despite using the highest available OSA resolution, it is not high enough to see the peaks of individual modes. Therefore, we used measurements of interference through a double-slit plate or a diffuser in the following to properly evaluate the coherence of the VCSELs.

To test the spatial coherence of the VCSELs, we used the double-slit experiment setup shown in Fig. 3(a). The spatial coherence of the emitted light causes the formation of interference patterns known as fringes. Each slit has a width, w, of 50 μm, and the distance between them, S, is 500 μm. We can then calculate the expected period of the fringes, P, formed on the plane of the camera using
P=λD/S,
(1)
where D is the distance between the double-slit plate and the camera (11 cm). The period is then estimated to be around 207 μm, which matches that seen in the fringes captured by the camera, as shown in Figs. 3(c) and 3(d). This verifies that the observed fringes are because of the interference caused by the two slits.
FIG. 3.

(a) The double-slit experiment setup used to characterize the spatial coherence of the VCSELs. The neutral-density (ND) filter is used to avoid saturating the camera. (b) The visibility of the fringes captured using the charge-coupled device (CCD) camera on the left hand side after the light is collimated using an objective and directed to the double-slit plate using the shortpass dichroic mirror, whose cutoff wavelength is around 650 nm. The fringes from the conventional VCSEL and the chaotic-cavity VCSEL are shown in (c) and (d), respectively.39 The stripes consist of 51 × 1440 pixels2 taken around the maximum intensity at each injection current. The dashed gray boxes mark the maximum captured intensity.

FIG. 3.

(a) The double-slit experiment setup used to characterize the spatial coherence of the VCSELs. The neutral-density (ND) filter is used to avoid saturating the camera. (b) The visibility of the fringes captured using the charge-coupled device (CCD) camera on the left hand side after the light is collimated using an objective and directed to the double-slit plate using the shortpass dichroic mirror, whose cutoff wavelength is around 650 nm. The fringes from the conventional VCSEL and the chaotic-cavity VCSEL are shown in (c) and (d), respectively.39 The stripes consist of 51 × 1440 pixels2 taken around the maximum intensity at each injection current. The dashed gray boxes mark the maximum captured intensity.

Close modal
The visibility of the fringes depends on the degree of spatial coherence. Therefore, a lower visibility is expected if there is an increase in the number of mutually incoherent modes emitted from the VCSEL. The visibility values for a conventional VCSEL and a chaotic-cavity VCSEL are shown in Fig. 3(b) at different injection currents above threshold. The visibility, V, is calculated using
V=ImaxIminImax+Imin,
(2)
where Imax and Imin are the intensity of the central peak and its adjacent trough, respectively. As the current increases, more modes start to lase, lowering the fringe visibility. However, at a certain point, the efficiency starts to drop due to heating, and fewer modes are emitted. The dashed box marks the values at the maximum captured intensity from the camera. The visibility is calculated from 51-pixel stripes from the camera around the maximum intensity point, which are shown in Figs. 3(c) and 3(d) for the conventional and chaotic-cavity VCSELs, respectively. The current values are labeled on each stripe. From these measurements, the D-shaped VCSEL exhibits significantly lower spatial coherence, with a visibility around 0.1 around its maximum intensity (compared to 0.38 from the conventional VCSEL).

After verifying the lower coherence of the chaotic-cavity VCSEL, which can help in improving its illumination performance, we tested its performance in an OWC link using the setup in Fig. 4(a). The tested VCSELs are probed using a ground-signal (GS) probe with the help of the top charge-coupled device (CCD) camera. The stage temperature is set to 17 °C using a thermoelectric cooler (TEC). The light is collected using a 50× objective lens and reflected through a 650-nm shortpass dichroic mirror to the 10-GHz amplified photodetector (PD). The light is focused on the active area using a 20× objective lens. The alignment CCD camera is used to align the beam spot on the PD (which is mounted on a three-dimensional translation stage). A variable optical attenuator (VOA) is used to avoid saturation of the PD. The DC port of the PD is used for alignment, whereas the received signal is collected from the AC port.

FIG. 4.

(a) The experimental setup used to test the communication performance of the VCSELs. The top charge-coupled device (CCD) camera is used for probing, while the one on the left is used for aligning the light with the active area of the amplified photodetector (PD). The variable optical attenuator (VOA) is used to make sure that the received optical power is within the linear dynamic range of the PD. The light is collimated using a 50× objective (obj.) and focused on the PD using a 20× obj. A thermoelectric cooler (TEC) is used to set the temperature of the stage at 17 °C. (b) and (c) Show the normalized frequency response of 6 O-shaped and 9 D-shaped VCSELs, respectively. (d) The bit error ratio (BER) of the received on-off keying (OOK) signals using a D-shaped VCSEL. The BER values are averaged over a period of 10 s. The dashed line shows the 7% overhead forward error correction (FEC) limit. (e) The eye diagram of the received signal with a data rate of 5 GB/s.

FIG. 4.

(a) The experimental setup used to test the communication performance of the VCSELs. The top charge-coupled device (CCD) camera is used for probing, while the one on the left is used for aligning the light with the active area of the amplified photodetector (PD). The variable optical attenuator (VOA) is used to make sure that the received optical power is within the linear dynamic range of the PD. The light is collimated using a 50× objective (obj.) and focused on the PD using a 20× obj. A thermoelectric cooler (TEC) is used to set the temperature of the stage at 17 °C. (b) and (c) Show the normalized frequency response of 6 O-shaped and 9 D-shaped VCSELs, respectively. (d) The bit error ratio (BER) of the received on-off keying (OOK) signals using a D-shaped VCSEL. The BER values are averaged over a period of 10 s. The dashed line shows the 7% overhead forward error correction (FEC) limit. (e) The eye diagram of the received signal with a data rate of 5 GB/s.

Close modal

We tested the frequency response of the two different types of VCSELs, as shown in Figs. 4(b) and 4(c). The AC signal is set to −17 dBm, the intermediate frequency (IF) bandwidth is set to 35 kHz, and the response is averaged over 30 scans. We overlay the normalized frequency response plots of VCSELs of the same shape in the same figure. These plots are recorded around the injection current at which the widest response for each VCSEL is observed. For the conventional VCSELs, the maximum observed −3 dB bandwidth (marked by the gray dashed lines) is around 5.6 GHz, whereas for the chaotic-cavity VCSELs it is around 5.1 GHz. From these results, it is shown that both types of VCSELs have similar high-speed performance. More importantly, these low-coherent broad-area VCSELs can support higher speeds as compared to the fastest reported SLDs (2.5 GHz).33 

We then tested the communication speed of the chaotic-cavity VCSEL using on-off keying (OOK). The injection current of the VCSEL was set to 90 mA. The AC signal is a 223 − 1 pseudorandom binary sequence (PRBS) generated by a bit error ratio (BER) tester. The signal is set to 500 mV peak-to-peak and passes through a fixed 4-GHz amplifier and a variable attenuator to control its amplitude. The signal is combined with the DC bias by a bias-tee, which in turn is connected to the GS probe, and the received signal is fed back to the BER tester. The BER values recorded over a period of 10 s at different data rates are shown in Fig. 4(d). The highest data rate achievable with a BER below the 7% overhead forward error correction (FEC) BER limit (3.8 × 10−3) is 5.5 GB/s. This BER value is used as a benchmark to compare with other studies in the literature that use the same value. The maximum net data rate using OOK at this BER value, taking the 7% overhead into account, is then around 5.1 GB/s. Figure 4(e) shows an open-eye diagram at a data rate of 5 GB/s. For applications that require a more stable communication link, the data rate can be reduced to achieve the required BER [as shown in Fig. 4(d)].

To improve the data rate even further, orthogonal frequency-division multiplexing (OFDM) was used instead of OOK. Given the flexibility of OFDM, the transmitted signal’s frequency response can be designed to make full use of the frequency response of the communication system. To implement it, the signals are generated offline and fed to an arbitrary waveform generator (AWG), whose AC peak-to-peak amplitude was set to 250 mV. The sampling rate was set to 8 × 109 samples/s. The received signal is recorded using an oscilloscope at a sampling rate of 50 × 109 samples/s. The transmitted signal consisted of 150 OFDM symbols. The fast Fourier transform for the OFDM signal was set to 1024, and the first five subcarriers (39 MHz) were left unused to avoid the low response of the used electronics in the low-frequency range. Data were loaded on the subsequent 500 subcarriers (0.039–3.9 GHz). The block diagram of the signal processing steps is shown in Fig. 5(a). More details are in the supplementary material and Fig. S2.

FIG. 5.

(a) The block diagram of the generation and detection of orthogonal frequency-division multiplexing (OFDM) signals. (b) The maximum and used spectral efficiency values. (c) The signal-to-noise ratio (SNR) and the power loading factor used to adjust the power assigned to each subcarrier. (d) The constellation diagrams of the different QAM orders from the received signal. All symbols from subcarriers with the same QAM order are superimposed in one diagram.

FIG. 5.

(a) The block diagram of the generation and detection of orthogonal frequency-division multiplexing (OFDM) signals. (b) The maximum and used spectral efficiency values. (c) The signal-to-noise ratio (SNR) and the power loading factor used to adjust the power assigned to each subcarrier. (d) The constellation diagrams of the different QAM orders from the received signal. All symbols from subcarriers with the same QAM order are superimposed in one diagram.

Close modal
First, we load 4-quadrature amplitude modulation (4-QAM) symbols on all subcarriers. The received signal is then processed, and the signal-to-noise ratio (SNR) of each subcarrier is calculated from the error vector magnitude (EVM) averaged from all 150 OFDM symbols. The SNR vs subcarrier plot is shown in Fig. 5(c). From the SNR, the maximum spectral efficiency, SE, for each subcarrier is calculated using
SE=log2(1+SNR).
(3)
Based on the spectral efficiency, different QAM orders up to 128-QAM are assigned to each subcarrier. The maximum and used spectral efficiency values per subcarrier are shown in Fig. 5(b). The gross data rate from this bit loading scheme is 14.1 GB/s. Moreover, the SNR is also used to tune the power sent through each subcarrier by multiplying it by a factor, as shown in Fig. 5(c). The received signal based on this adaptive bit/power loading scheme is then processed, and the BER is calculated. Figure 5(d) shows the constellation diagrams of all QAM orders used. To achieve a BER below the 7% overhead FEC limit, six out of the 150 OFDM symbols are used as training symbols for synchronization and channel estimation in the post-equalization process. The BER in that case is around (3.7 × 10−3), and the net data rate, taking into account the 7% overhead and the training symbols, is around 12.6 GB/s (see Fig. S3 in the supplementary material for a comparison with OOK). It is worth mentioning that, just like any communication system, lower BER values are achievable by simply lowering the data rate. The BER value chosen here is used as a benchmark in most OWC works in the literature, making the comparison between systems easier. The achieved net data rate using the chaotic-cavity VCSEL (12.6 GB/s) far exceeds the record data rate achieved using SLDs (4.6 GB/s)34 while still maintaining a lower speckle contrast (0.1 compared to 0.2).30,32

Although the spatial coherence and the speckle contrast have been shown to be lower for chaotic-cavity VCSELs, the speckle contrast is still higher than the threshold for human perception (0.03).15 The speckle contrast is defined as the standard deviation of the speckle patterns formed by passing the light through a diffuser after normalizing its mean. The number of modes is estimated by taking the reciprocal of the square of the speckle contrast. To lower it further, it is possible to increase the number of emitted modes using an array of chaotic-cavity VCSELs, but this requires more power and a larger chip size. It was also shown that pulsed operation can lower the spatial coherence,1,36–38 but this is not suitable for applications that need CW operation, such as communication. Instead, we show that the communication signal can be designed to lower the apparent spatial coherence. First, we test the coherence of the conventional circular-cavity VCSEL by measuring the speckle contrast formed after passing the beam through a diffuser. The measurement is performed with no AC signal added, with an OFDM signal, and with an OFDM signal with a short chirp added in between symbols. The chirp signal is a sine wave whose frequency changes linearly in time from 0 to 500 MHz. It is defined as a function of time, t as sin(500 × 106πt2/T), where T is the duration of the chirp signal, which is designed to be a fifth of the entire AC signal. In other words, 20% of the duration of the signal is used to transmit the chirp signal, which lowers the net data rate by 20%.

Figures 6(a)6(c) show the AC signal used in each scenario, whereas Figs. 6(d)6(f) show the corresponding speckle patterns and the estimated number of modes from each of them. Since the beam does not carry uniform power across the recorded space, we normalize each speckle pattern by dividing it by a smoothed version of the image using a 500 × 500 moving mean filter. We then calculate the speckle contrast and the estimated number of modes. The values also closely match the values calculated from 501 × 501 cropped versions of the image without normalization. In the case where no AC signal was transmitted, the estimated number of modes was around 88. Adding the OFDM signal slightly increases that to 131. However, by adding the chirp signal, the estimated number of modes is increased to 196, which is more than double the value with no added signals. Taking into account the 20% drop due to the added chirp signal and using 15 symbols for training, the net data rate in this case is around 4.8 GB/s, and the BER is 3.7 × 10−3 using bit and power loading (see the supplementary materials and Fig. S4 for more details). The histograms of the recorded speckle patterns after normalizing by their respective means are shown in Fig. 6(g). The narrower histograms indicate a lower speckle density in the image.

FIG. 6.

(a)–(c) Respectively show the used AC signals for the speckle pattern measurements, whereas (d)–(f) show the corresponding speckle patterns. (g) The corresponding probability density functions of the intensity normalized by the mean obtained from the calculated histograms.

FIG. 6.

(a)–(c) Respectively show the used AC signals for the speckle pattern measurements, whereas (d)–(f) show the corresponding speckle patterns. (g) The corresponding probability density functions of the intensity normalized by the mean obtained from the calculated histograms.

Close modal

We used the same technique on a chaotic-cavity VCSEL. Figures 7(a)7(c), respectively, show the used AC signal, the speckle pattern, and the near-filed profile of the chaotic-cavity VCSEL when the AC signal is off. Figures 7(d)7(f) show them when the OFDM signal is combined with a chirp signal. Figure 7(g) shows the histograms. The speckle pattern becomes dramatically smoother when OFDM and chirp signals are used. The number of modes increases from around 131 to 454, achieving a speckle contrast of 4.7%. That means that using only two chaotic-cavity VCSELs with this technique can achieve almost speckle-free illumination. The near-field profiles show that the designed AC signal results in a more uniform optical field. This is most likely due to the fast-changing AC signal modifying the supported transverse modes. The exposure time of the camera (30 ms) is not short enough to capture the changes, and the result is a smooth image. The same applies to the speckle patterns. In fact, this is similar to reducing the speckle contrast using a rotating diffuser but without resorting to using moving parts, which makes this technique more compact, significantly easier to use, and more reliable.

FIG. 7.

(a)–(c) Show the AC signal, the speckle pattern, and the near-field profile of the chaotic cavity VCSEL. (d)–(f) Show them with an OFDM and a chirp signal are combined. (g) The probability density functions of the intensity normalized by the respective means.

FIG. 7.

(a)–(c) Show the AC signal, the speckle pattern, and the near-field profile of the chaotic cavity VCSEL. (d)–(f) Show them with an OFDM and a chirp signal are combined. (g) The probability density functions of the intensity normalized by the respective means.

Close modal

In this work, we achieved low coherence surface emission by modifying the cavity shape of VCSELs and carefully designing the injected signal. We first showed the effects of changing the cavity shape on the spatial coherence by showing a significant decrease in the fringe visibility in a double-slit experiment. We then showed that these chaotic-cavity VCSELs can support high speed communication with a modulation bandwidth of up to 5 GHz and an achievable net data rate of 12.6 GB/s using OFDM. We then demonstrated a new technique for lowering the apparent spatial coherence of these VCSELs by using a chirp signal in between OFDM symbols, resulting in an increase in the number of modes up to 454 transverse modes. Compared to conventional broad-area VCSELs with slightly larger areas, these two combined techniques result in a fivefold increase in the number of modes. More interestingly, the modulation bandwidth of the chaotic-cavity VCSEL is double that of the record bandwidth of SLDs (2.5 GHz) with even higher optical power. It also achieved a much higher data rate (12.6 GB/s as compared to 4.6 GB/s using similar modulation techniques). The results presented here highlight the potential of low-coherence surface-emitting lasers (LCSELs) in simultaneous illumination and OWC, especially when the techniques reported here are implemented on visible-light VCSELs.

Additional improvements to performance are possible. Optimizing the chirp signal in terms of duration, initial and final frequencies, and amplitude can help lower the speckle contrast. They can also be tuned to maintain a higher data rate, depending on the application. Moreover, the techniques presented here are not exclusive to near-infrared VCSELs. The present work paves the way toward the implementation of visible-light VCSELs, which are of high interest for indoor illumination and Li-Fi systems. The shorter wavelength of visible light makes it easier to support even more modes with lower diffraction losses in the same VCSEL size.

The supplementary material includes details on the VCSEL wafer, the electrical and optical characterization, and the communication scheme.

Huawei Technologies Co., Ltd. (Grant No. YBN2020085017); King Abdullah University of Science and Technology (Grant Nos. BAS/1/1614-01-01, RGC/3/4275-01-01, GEN/1/6607-01-01, KCR/1/2081-01-01, and KCR/1/4114-01-01). T.K.N. and B.S.O. acknowledge funding from King Abdullah University of Science and Technology (KAUST) Research Funding (KRF) under Award No. ORA-2022-5313.

The authors have no conflicts to disclose.

Omar Alkhazragi: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Funding acquisition (supporting); Investigation (lead); Methodology (lead); Software (lead); Validation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (equal). Ming Dong: Data curation (supporting); Software (supporting); Writing – review & editing (supporting). Liang Chen: Conceptualization (supporting); Software (supporting). Meiwei Kong: Software (supporting). Georgian Melinte: Data curation (supporting); Writing – review & editing (supporting). Dong Liang: Investigation (supporting); Writing – review & editing (supporting). Tien Khee Ng: Conceptualization (supporting); Funding acquisition (supporting); Investigation (supporting); Project administration (equal); Resources (supporting); Supervision (supporting); Writing – review & editing (supporting). Junping Zhang: Project administration (equal); Supervision (equal); Writing – review & editing (equal). Hakan Bagci: Conceptualization (equal); Investigation (equal); Software (equal); Supervision (equal); Writing – review & editing (equal). Boon S. Ooi: Conceptualization (equal); Funding acquisition (equal); Investigation (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
H.
Cao
,
R.
Chriki
,
S.
Bittner
,
A. A.
Friesem
, and
N.
Davidson
, “
Complex lasers with controllable coherence
,”
Nat. Rev. Phys.
1
,
156
168
(
2019
).
2.
B.
Redding
,
M. A.
Choma
, and
H.
Cao
, “
Speckle-free laser imaging using random laser illumination
,”
Nat. Photonics
6
,
355
359
(
2012
).
3.
Y.
Peng
,
S.
Choi
,
J.
Kim
, and
G.
Wetzstein
, “
Speckle-free holography with partially coherent light sources and camera-in-the-loop calibration
,”
Sci. Adv.
7
,
eabg5040
(
2021
).
4.
O.
Alkhazragi
,
J. A.
Holguín-Lerma
,
T. K.
Ng
, and
B. S.
Ooi
, “
Visible-light laser diodes and superluminescent diodes: Characteristics and applications
,” in digital
Encyclopedia of Applied Physics
(Wiley-VCH Verlag GmbH & Co.,
2021
).
5.
J. P.
Reithmaier
,
G.
Eisenstein
, and
A.
Forchel
, “
InAs/InP quantum-dash lasers and amplifiers
,”
Proc. IEEE
95
,
1779
1790
(
2007
).
6.
M. Z. M.
Khan
,
T. K.
Ng
,
C.-S.
Lee
,
P.
Bhattacharya
, and
B. S.
Ooi
, “
Investigation of chirped InAs/InGaAlAs/InP quantum dash lasers as broadband emitters
,”
IEEE J. Quantum Electron.
50
,
51
61
(
2014
).
7.
B. S.
Ooi
,
H.
Susanto Djie
,
Y.
Wang
,
C. L.
Tan
,
J. C. M.
Hwang
,
X. M.
Fang
,
J. M.
Fastenau
,
A. W. K.
Liu
,
G. T.
Dang
, and
W. H.
Chang
, “
Quantum dashes on InP substrate for broadband emitter applications
,”
IEEE J. Sel. Top. Quantum Electron.
14
,
1230
1238
(
2008
).
8.
Z.
Mi
,
J.
Yang
, and
P.
Bhattacharya
, “
Growth and characteristics of p-doped InAs tunnel injection quantum-dash lasers on InP
,”
IEEE Photonics Technol. Lett.
18
,
1377
1379
(
2006
).
9.
E. V.
Andreeva
,
A. S.
Anikeev
,
S. N.
Il’chenko
,
A.
Chamorovskiy
,
V. R.
Shidlovski
, and
S. D.
Yakubovich
, “
Highly efficient superluminescent diodes and SLD-based combined light sources of red spectral range for applications in biomedical imaging
,”
Proc. SPIE
10483
,
104832T
(
2018
).
10.
A. A.
Alatawi
,
J. A.
Holguin-Lerma
,
C. H.
Kang
,
C.
Shen
,
R. C.
Subedi
,
A. M.
Albadri
,
A. Y.
Alyamani
,
T. K.
Ng
, and
B. S.
Ooi
, “
High-power blue superluminescent diode for high CRI lighting and high-speed visible light communication
,”
Opt. Express
26
,
26355
26364
(
2018
).
11.
C.
Shen
,
J. A.
Holguin-Lerma
,
A. A.
Alatawi
,
P.
Zou
,
N.
Chi
,
T. K.
Ng
, and
B. S.
Ooi
, “
Group-III-nitride superluminescent diodes for solid-state lighting and high-speed visible light communications
,”
IEEE J. Sel. Top. Quantum Electron.
25
,
2000110
(
2019
).
12.
A.
Kafar
,
S.
Stanczyk
,
D.
Schiavon
,
T.
Suski
, and
P.
Perlin
, “
Review—Review on optimization and current status of (Al, In)GaN superluminescent diodes
,”
ECS J. Solid State Sci. Technol.
9
,
015010
(
2019
).
13.
C. K.
Hitzenberger
,
M.
Danner
,
W.
Drexler
, and
A. F.
Fercher
, “
Measurement of the spatial coherence of superluminescent diodes
,”
J. Mod. Opt.
46
,
1763
1774
(
1999
).
14.
J.
Yang
,
J.
Heo
,
T.
Zhu
,
J.
Xu
,
J.
Topolancik
,
F.
Vollmer
,
R.
Ilic
, and
P.
Bhattacharya
, “
Enhanced photoluminescence from embedded PbSe colloidal quantum dots in silicon-based random photonic crystal microcavities
,”
Appl. Phys. Lett.
92
,
261110
(
2008
).
15.
B.
Redding
,
A.
Cerjan
,
X.
Huang
,
M. L.
Lee
,
A. D.
Stone
,
M. A.
Choma
, and
H.
Cao
, “
Low spatial coherence electrically pumped semiconductor laser for speckle-free full-field imaging
,”
Proc. Natl. Acad. Sci. U. S. A.
112
,
1304
1309
(
2015
).
16.
A.
Consoli
,
N.
Caselli
, and
C.
López
, “
Electrically driven random lasing from a modified Fabry–Pérot laser diode
,”
Nat. Photonics
16
,
219
225
(
2022
).
17.
H.
Soda
,
K. . i.
Iga
,
C.
Kitahara
, and
Y.
Suematsu
, “
GaInAsP/InP surface emitting injection lasers
,”
Jpn. J. Appl. Phys.
18
,
2329
(
1979
).
18.
K.
Iga
, “
Forty years of vertical-cavity surface-emitting laser: Invention and innovation
,”
Jpn. J. Appl. Phys.
57
,
08PA01
(
2018
).
19.
R.
Michalzik
, in
VCSELs: Fundamentals, Technology and Applications of Vertical-Cavity Surface-Emitting Lasers
(
Springer
,
Berlin, Heidelberg
,
2012
), Chap. 1.
20.
B. D.
Padullaparthi
,
J. A.
Tatum
, and
K.
Iga
, “
VCSEL industry: Communication and sensing
,” in
VCSEL Industry
(
John Wiley & Sons
,
Incorporated, Hoboken
,
2021
), Chap. 1.
21.
M.
Kuramoto
,
S.
Kobayashi
,
T.
Akagi
,
K.
Tazawa
,
K.
Tanaka
,
K.
Nakata
, and
T.
Saito
, “
Watt-class blue vertical-cavity surface-emitting laser arrays
,”
Appl. Phys. Express
12
,
091004
(
2019
).
22.
K.
Terao
,
H.
Nagai
,
D.
Morita
,
S.
Masui
,
T.
Yanamoto
, and
S.
ichi Nagahama
, “
Blue and green GaN-based vertical-cavity surface-emitting lasers with AlInN/GaN DBR
,”
Proc. SPIE
11686
,
116860E
(
2021
).
23.
R. T.
ElAfandy
,
J.-H.
Kang
,
B.
Li
,
T. K.
Kim
,
J. S.
Kwak
, and
J.
Han
, “
Room-temperature operation of c-plane gan vertical cavity surface emitting laser on conductive nanoporous distributed Bragg reflector
,”
Appl. Phys. Lett.
117
,
011101
(
2020
).
24.
H.
Haas
,
L.
Yin
,
Y.
Wang
, and
C.
Chen
, “
What is LiFi?
,”
J. Lightwave Technol.
34
,
1533
1544
(
2016
).
25.
H.
Haas
, “
LiFi is a paradigm-shifting 5G technology
,”
Rev. Phys.
3
,
26
31
(
2018
).
26.
I.
Dursun
,
C.
Shen
,
M. R.
Parida
,
J.
Pan
,
S. P.
Sarmah
,
D.
Priante
,
N.
Alyami
,
J.
Liu
,
M. I.
Saidaminov
,
M. S.
Alias
,
A. L.
Abdelhady
,
T. K.
Ng
,
O. F.
Mohammed
,
B. S.
Ooi
, and
O. M.
Bakr
, “
Perovskite nanocrystals as a color converter for visible light communication
,”
ACS Photonics
3
,
1150
1156
(
2016
).
27.
M.
Ayyash
,
H.
Elgala
,
A.
Khreishah
,
V.
Jungnickel
,
T.
Little
,
S.
Shao
,
M.
Rahaim
,
D.
Schulz
,
J.
Hilt
, and
R.
Freund
, “
Coexistence of WiFi and LiFi toward 5G: Concepts, opportunities, and challenges
,”
IEEE Commun. Mag.
54
,
64
71
(
2016
).
28.
Z.
Chen
,
D. A.
Basnayaka
, and
H.
Haas
, “
Space division multiple access for optical attocell network using angle diversity transmitters
,”
J. Lightwave Technol.
35
,
2118
2131
(
2017
).
29.
E.
Sarbazi
,
H.
Kazemi
,
M. D.
Soltani
,
M.
Safari
, and
H.
Haas
, “
A Tb/s indoor optical wireless access system using VCSEL arrays
,” in
2020 IEEE 31st Annual International Symposium on Personal, Indoor and Mobile Radio Communications
(
IEEE
,
2020
), pp.
1
6
.
30.
O.
Alkhazragi
,
M.
Dong
,
L.
Chen
,
D.
Liang
,
T. K.
Ng
,
J.
Zhang
,
H.
Bagci
, and
B. S.
Ooi
, “
Modifying the coherence of vertical-cavity surface-emitting lasers using chaotic cavities
,”
Optica
10
,
191
199
(
2023
).
31.
D.
Inoue
,
R.
Kubota
,
T.
Aoki
,
T.
Ishizuka
,
M.
Yanagisawa
, and
H.
Shoji
, “
1× 4 VCSEL arrays with uniform spectral and noise properties by using rotationally asymmetric oxide aperture for 400 Gbit/s applications
,”
Proc. SPIE
11300
,
113000F
(
2020
).
32.
B.
Redding
,
P.
Ahmadi
,
V.
Mokan
,
M.
Seifert
,
M. A.
Choma
, and
H.
Cao
, “
Low-spatial-coherence high-radiance broadband fiber source for speckle free imaging
,”
Opt. Lett.
40
,
4607
4610
(
2015
).
33.
A.
Rashidi
,
A. K.
Rishinaramangalam
,
A. A.
Aragon
,
S.
Mishkat-Ul-Masabih
,
M.
Monavarian
,
C.
Lee
,
S. P.
Denbaars
, and
D. F.
Feezell
, “
High-speed nonpolar InGaN/GaN superluminescent diode with 2.5 GHz modulation bandwidth
,”
IEEE Photonics Technol. Lett.
32
,
383
386
(
2020
).
34.
D.
Li
,
C.
Ma
,
J.
Wang
,
F.
Hu
,
Y.
Hou
,
S.
Wang
,
J.
Hu
,
S.
Yi
,
Y.
Ma
,
J.
Shi
,
J.
Zhang
,
Z.
Li
,
N.
Chi
,
L.
Xia
, and
C.
Shen
, “
High-speed GaN-based superluminescent diode for 4.57 Gbps visible light communication
,”
Crystals
12
,
191
(
2022
).
35.
A.
Brejnak
,
M.
Gębski
,
A. K.
Sokół
,
M.
Marciniak
,
M.
Wasiak
,
J.
Muszalski
,
J. A.
Lott
,
I.
Fischer
, and
T.
Czyszanowski
, “
Boosting the output power of large-aperture lasers by breaking their circular symmetry
,”
Optica
8
,
1167
1175
(
2021
).
36.
S. K.
Mandre
,
W.
Elsäßer
,
I.
Fischer
,
M.
Peeters
, and
G.
Verschaffelt
, “
Evolution from modal to spatially incoherent emission of a broad-area VCSEL
,”
Opt. Express
16
,
4452
4464
(
2008
).
37.
M.
Peeters
,
G.
Verschaffelt
,
H.
Thienpont
,
S. K.
Mandre
,
I.
Fischer
, and
M.
Grabherr
, “
Spatial decoherence of pulsed broad-area vertical-cavity surface-emitting lasers
,”
Opt. Express
13
,
9337
9345
(
2005
).
38.
F.
Riechert
,
G.
Verschaffelt
,
M.
Peeters
,
G.
Bastian
,
U.
Lemmer
, and
I.
Fischer
, “
Speckle characteristics of a broad-area VCSEL in the incoherent emission regime
,”
Opt. Commun.
281
,
4424
4431
(
2008
).
39.
O.
Alkhazragi
,
M.
Dong
,
L.
Chen
,
D.
Liang
,
T. K.
Ng
,
J.
Zhang
,
H.
Bagci
, and
B. S.
Ooi
, “
Spatial Coherence of Chaotic-Cavity Surface-Emitting Lasers
,” in
CLEO 2023, Technical Digest Series
(Optica Publishing Group, in press).

Supplementary Material