Review of Laser Doping and its Applications in Silicon Solar Cells

Laser-doped selective emitter diffusion techniques have become mainstream in solar cell manufacture covering 60% of the market share in 2022 and are expected to continue to grow to above 90% within the next five years (ITRPV). This was a very rapid uptake of technology, coming from only ∼10% penetration in 2018, and has enabled over 20 fA/cm2 front recombination current reductions on the dominant passivated emitter and rear cell concepts in the same short period. In this article, a broad overview of key concepts in relation to laser doping methods relevant to solar cell manufacturing is given. We first discuss the basic mechanisms behind laser doping along with the benefits over conventional doping methods. The main laser doping approaches reported in the literature are then discussed, along with implications for metallization strategy, particularly in relation to selective emitter and back surface field formation in the dominant passivated emitter and rear cell technology. Different cell concepts that have benefited from the application of laser doping are also discussed. In the last section, we discuss the main defects induced by laser processing of silicon which affect the finished devices, potential and debated causes, as well as some commonly applied treatments for their mitigation.


I. INTRODUCTION
M ODERN laser technology has developed significantly in the past few decades. Fast advances in the understanding and manufacture of lasers have made possible the mass production of a wide variety of high-efficiency light sources in a myriad of wavelengths, pulse shapes/widths, as well as compactness. From an industrial perspective, lasers offer a precise, reliable, and cost-effective substitute to conventionally used processing tools. In the push to higher efficiency industrial silicon solar cells, laser doping has been a particularly important application of laser technology that enables the formation of localized heavily doped regions on a silicon surface, with its main advantages being the following. 1) Room temperature processing, eliminating the need for high-temperature furnaces and so reducing processing etching. The superiority of laser doping over conventional hightemperature tube furnace annealing has been demonstrated abundantly [1], [2], [3], [4], [5], [6], [7], [8], [9].
Laser doping has become the standard process for forming selective emitters on the front surface of the dominant passivated emitter and rear cell (PERC) technology and can also be used as an elegant process to form the heavily doped p-type back surface fields on their rear surface [10]. It has the potential to also be applied to form the heavily doped contact regions in other high efficiency solar cell technologies, such as n-PERT (passivated emitter rear totally diffused), tunnel-oxide passivated contact (TOPCon) and interdigitated back contact (IBC) cells to name just a few [11].
The principle of laser doping results from a huge power transfer that occurs when a concentrated light beam is focused onto a silicon crystal bringing about a nearly immediate raise of lattice temperature (tens of ns). This temperature can reach levels over 1414°C which is followed by a likewise fast quenching process (∼735°C/s) [11]. Such a process, when adequately high energies are utilized, results in the simultaneous rapid melting as well as recrystallisation of laser-subjected locations in the Si crystals, with the crucial benefit of a "near-zero" thermal budget in the rest of the volume. Furthermore, when laser doping is performed in the liquified state, the diffusion coefficient of dopants is orders of magnitude greater than in the solid-state, making dopant diffusion occur in a split second compared to the several-hour processes required using conventional thermal diffusion processes. Manipulation of the laser fluence, wavelength as well as pulse size throughout the doping step can be utilized to finely tune the width and depth of dissolved regions in any type of provided sample for a reliable diffusion or activation of dopants.
The aim of this review article is to provide a broad overview of the laser doping process and how it has been, and can potentially be utilized in silicon solar cells. It will first provide a brief overview on the basic mechanisms of the process, highlighting to the differences and advantages over conventional doping techniques. It will then review applications of laser doping in silicon solar cells, including the most relevant use cases of This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ selective emitter and back surface field formation, with emphasis on the industrially dominant PERC cell. Most common dopant sources that have been used for laser doping of solar cells in the literature will be summarized. Specific applications including laser doped semiconductor finger cells, interdigitated back contact cells and potential applications in industrial TOPCon cells will be discussed. Finally, the issue of laser-induced defects and methods to avoid or passivate them will be covered.

II. CONVENTIONAL DOPING AND LASER DOPING
Conventional techniques for doping of semiconductors such as ion implantation and thermal (gas phase) diffusion have been largely studied throughout the years and are mature technologies. On one hand, ion implantation, the capacity of which was first recognized in 1954, has become one of the most crucial technologies for today's integrated circuit manufacture. Ion implantation has enabled the formation of precise, reliable as well as repeatable doping profiles in semiconductors, along with high tunability in terms of doping focus and depth. Nonetheless, obstacles imposed by the exact control of low-energy ion beams needed for shallow and also sharp joint formation, have represented a considerable disadvantage for its application [12]. Amongst them, damage brought on by very energetic dopants bombardment of the silicon surface which requires additional thermal processing for its elimination as well as dopant activation might result in precipitation, unwanted dopant redistribution, wafer contamination and also surface degradation [13]. The combination of these aspects has hence hampered the advancement of this technology for mainstream solar cell manufacture which overall likewise calls for low cost and high throughput processes.
Thermal diffusions, on the other hand, have been the industrial standard for emitter formation in silicon solar cells since the 1970s. In particular, using phosphorus oxychloride (POCl 3 ) diffusion for electron collector formation on p-type silicon wafers and boron tribromide (BBr 3 ) on n-type silicon wafers. Both of which have been dominating the market as a result of their low costs, stability and also high throughput, especially in the case of POCl 3 diffusions [14]. Despite these perks, optimal thermal diffusions require some rather finetuning of the process and present a series of challenges and drawbacks which need to be addressed.
Thermal diffusion processes begin with the exposure of the silicon wafers to a precursor gas at high temperatures. This usually requires dedicated furnaces to avoid cross-contamination and results in the growth of a phosphorus-rich or boron-rich layer at the silicon surfaces, commonly referred to as phosphosilicate glass (PSG) and borosilicate glass (BSG), respectively. Depending on the dopant species, high-temperature "drive-in" processes are utilized after the dopant layer growth, which enables the diffusion of dopants into the silicon substrate in the solid state. As a result of the differences in diffusion coefficients from the various dopant impurities, the length and temperatures used for junction formation change. A clear example would be that of a boron (B) diffusion, which requires higher temperatures and longer processes when compared to a phosphorus (P) diffusion.
A typical B diffusion will take more than 80 mins for junction formation at a temperature of 900-1000°C [15], whereas P diffusions on the other hand are usually executed at temperatures of 800-900°C for approximately 50 mins [16]. During the drive-in step, the rich layer dopants present at the surface reach a point of dopant mobility which allow their diffusion into the silicon subsurface. Some of these impurities, once in the silicon crystal, then hop around the silicon atoms and occupy substitutional lattice sites where they become electrically active. Others, however, will instead diffuse to form dopant clusters, precipitates or inhabit interstitial sites, which can bring about severe recombination losses and decreased collection effectiveness of the photogenerated carriers in the so-called "dead layer" [17]. Such dopants are inactive and do not add to the material's conductivity. Process optimization methods such as reduction of POCl 3 circulation rates must then be implemented to decrease the concentration of nonactive P, even though a marginal concentration of these impurities is also needed for developing low resistivity contacts making it a sensitive process [18].
Another relevant consideration for thermal diffusions of phosphorus or boron into silicon wafers is the fact that it creates stress on the silicon lattice as an outcome of atomic distance inequality [19]. If not meticulously developed and managed, these stresses can surpass the crucial resolved shear tension, causing the emergence of intrinsic point defect, i.e., diffusion-induced dislocations and vacancies. In addition, thermal diffusion with POCl 3 gas inevitably results in huge thermal budgets and the diffusion of dopants across all surfaces of the wafer that are exposed to the gas in the furnace. The former might lead to undesired defect activation while the latter can cause the formation of a diffused region all around the wafer which effectively creates a low resistance path for photogenerated carriers from the front to the rear of the cell, likewise, drastically impacting its performance. Some applied strategies for the isolation consist of additional processing steps such as sawing, grinding, plasma etching, laser ablation, and rear wet etching [20], [21].
Very conveniently, laser doping has emerged to offer a much more flexible approach that overcomes several of the drawbacks presented by conventional diffusion. Even though such flexibility comes at the usual cost of finetuning the laser parameters for each application, they require only a fraction of the time and cost for its optimization compared to thermal diffusions as will be briefly discussed. Much of this optimization comes from the need for precise control of doping depths and shapes. In the case of deep junctions, the depth of the laser doped region heavily depends on the depth of the liquified area in the silicon, whereas the kind of laser used will define the dopant profile. The laser can either be pulsed, such as for Q-switched lasers, where the laser energy is rapidly pulsed, or continuous wave (CW). Pulsed lasers lead to much shallower melting zones typically around 1.5 µm, compared to the >10 µm for CW lasers before ablation, and so limiting the depth of the diffusion profile. In the case of short nano-second pulse-length lasers, the silicon commonly solidifies between successive pulses, meaning that the same quantity is re-molten with each pulse, resulting in a more uniform doping profile. Nevertheless, much longer (µs) pulse length lasers and CW lasers can lead to a much deeper injection of dopants, which can produce better doping profiles under the contacts and better field passivation at the metal/Si interface, as well as junction depths of up to 10 µm. Similar results can be achieved with mode-locked quasi-CW (Q-CW) lasers, which have much shorter (ps) pulse lengths but work at very high frequencies ranging from 80 to 120 MHz which are fast enough to prevent recrystallisation between successive pulses. Another important parameter is the wavelength of the laser. In the case of silicon, the absorption coefficient at 300 K varies by approximately five orders of magnitude across the absorption range, with the best absorption occurring at the lower end of the wavelength spectrum. Still, a wide range of lasers can be used for doping, typically used wavelengths include 355, 532, 1030, and 1064 nm. This is because silicon's absorption coefficient is largely temperature dependent and increases over an order of magnitude at most wavelengths at moderate temperatures (∼700°C) [22]. Thus, the rapid temperature increase of the silicon substrate during laser exposure accelerates the absorption changes in the whole absorption range, making it important to note that for typical Nd:YAG 532 nm lasers, this increase may lead to extremely shallow absorption and unwanted ablation.
Importantly, laser doping has allowed the easy implementation of selectively diffused surfaces which in the past were only possible using costly and complicated photolithography patterning processes which involved multiple high-temperature diffusion and oxidation steps, as well as many chemical etching steps. This has facilitated the use of lower quality substrates in higher efficiency cell architectures, such as multicrystalline silicon, which is known to degrade during very high-temperature processing [23], [24]. Deeper junctions are able to be formed much faster via laser doping than by conventional diffusion processes, owing to the higher temperatures that can be achieved locally using laser doping, and in many cases, the diffusion of dopants in the liquid state.

A. Current Industrial Approach
Current industrial cells opt for the approach of laser doping prior to dielectric deposition, combined with an aligned screen-printed fire-through metallization step [25]. A schematic of this structure is shown in Fig. 1. Laser doping is performed directly after the thermal diffusion step, utilizing either the PSG or BSG layer that forms during diffusion as a source of n-type or p-type dopants, respectively. A wide ∼100 µm strip of laser doping is required to facilitate the later alignment of the typical screen-printed metal fingers of 40-50 µm width. The frontal laser doping is performed in the solid state, to avoid melting of the textured pyramid surface and so ensure that the maximum optical absorption is maintained in the nonmetallized regions. After removal of the PSG or BSG layer and deposition of the passivating dielectric layer, an aligned screen-printing and co-firing process is performed. During firing, the metal penetrates through the dielectric layer to contact the underlying laser-doped region. An advantage of this approach is the more effective passivation of laser induced defects, as the passivating dielectric is deposited over the laser doped region. Additionally, laser-induced defects that form due to the thermal expansion mismatch of Si and the dielectric layer are avoided. Using this approach in conjunction with a fire-through metal contacting paste also means the need for a separate local contact opening step is avoided.

B. Laser Doping After Dielectric Deposition
The UNSW developed a laser doping selective emitter (LDSE) approach [27], later commercialized by Suntech Power as PLUTO [28], that involved lasing through a dielectric stack and melting and doping the underlying silicon. In this approach, the dopant source was typically externally deposited through either a spin-on-dopant (SOD) source or using an ultrasonic sprayer [29]. Doped silicon nanoparticle ink [30] or dielectric layers deposited via chemical vapor deposition containing dopant atoms such as phosphorous-doped SiN x , PSG, and BSG have also been used as the dopant source in a similar approach [29], [31], [32]. Some other viable dielectric stacks are discussed in Section IV. Alternatively, the doping may simply be achieved through redistribution of the dopants in the emitter itself [33]. The laser chemical processing (LCP) method developed by Fraunhofer ISE was typically carried out through a dielectric layer stack in a similar manner to the UNSW approach [34]. In this scenario, the laser was guided by a liquid jet containing a dopant source and so removing the need for a separate step for dopant deposition. One of the main advantages of said methods for laser doping postdielectric deposition is that, if combined with plating, self-aligned metallization is possible as the metal is only deposited on the exposed laser-doped regions. With this approach, laser doping coverage can be kept to a minimum of around 1%-2% (only the laser beam width) compared with the 10% required for alignment tolerance in the current industrial method using screen-printed contacts. The contribution of the laser-doped region to the total emitter recombination current density (J 0e ) is therefore reduced. Plating is also advantageous as it allows for very fine metallization widths (<20 µm), leading to lower shading losses and therefore facilitating more closely spaced fingers and higher sheet resistance emitters. Additionally, plated contacts have much higher conductivity and can achieve  lower contact resistance than screen-printed contacts, on the order of 0.1 mΩ·cm 2 [35]. A focused ion beam (FIB) image of a cross section of an LDSE cell with plated contacts is shown in Fig. 2 .
At present, however, the widespread commercial use of these self-aligned contacting schemes has been hindered by the challenges presented by this technology including: complexity of process optimization, contact adhesion, diffusion barriers, and waste management. In addition to these, the constant improvements of screen-printing technology such as the capability to form narrower fingers, alignment capabilities, and reduced Ag consumption have made the financial justification of shifting to Cu-based plated contacts more difficult. Therefore, screenprinting continues to be the main contact formation method for LDSE solar cells.

C. Simultaneous Laser Doping and Grooving for Laser Doped and Plated Buried Contacts
An alternative approach developed by Wang et al. [37] involved the formation of narrow (3-5 µm wide), deep (10-15 µm) and heavily doped grooves prior to dielectric deposition (see Fig. 3). Through fine optimization of the laser conditions, silicon is ablated to form the narrow groove whilst simultaneously heavily doping the groove walls with dopants from an SOD source. This approach capitalizes on the benefits of laser doping prior to dielectric deposition, but also facilitates an aligned contacting approach with plating due to the conformal nature of the plasma-enhanced chemical vapor deposition (PECVD) process which enables the groove to largely remain exposed for nucleation of plated metal. This technology achieved efficiencies over 19% for a device with full area aluminum back surface field, further improvements would be expected shifting to a PERC cell structure.

IV. LOCAL BACK SURFACE FIELD FORMATION USING DIELECTRIC FILMS AS THE DOPANT SOURCE
As it has been mentioned, localized doping of controlled dopant concentration and depth are required for advanced and high-efficiency solar cells. Early conventional approaches, therefore, required the application of laborious photolithography or patterning methods in combination with lengthy thermal diffusion sequences and their corresponding chemical removal steps. Furthermore, even when laser doping has been demonstrated as a much more flexible alternative, it still requires the presence of a dopant-containing film which in most cases is added during a previous fabrication step and also requires to be removed after laser processing. If however, the doping species are contained in a pre-existing (and ultimately required) dielectric stack, these extra steps can be eliminated. Such is the case of the standard dielectric stack which forms the rear side of a typical industrial PERC cell, comprising a stack of aluminum oxide (AlO x ) capped with SiN x . There, the AlO x is originally present as an excellent surface passivation layer due to its high density of fixed negative charge causing an accumulation effect on p-type silicon surfaces [38], [39]. However, and rather unintendedly, this same layer can act as a rich source of Al, a p-type dopant, for forming heavily doped p ++ regions in the underlying silicon using laser processing. This has now been demonstrated by several authors [40], [41], [42], [43], [44].
One challenge with laser doping from AlO x layers, however, is the limited solid solubility of aluminum in silicon above 1200°C (see Fig. 4). In contrast to boron dopants, the solid solubility of aluminum in silicon peaks at a temperature of ∼1100-1200°C (2 × 10 19 cm −3 ), and then reduces with increasing temperature. The peak active doping concentration using AlO x -based laser doping may therefore be lower than that when using boron dopants. Although, the use of Al dopants may help to avoid the formation of BO defects in the laser-doped region which is typically oxygen-rich [45].
Despite having a lower solubility limit, Martín et al. demonstrated a p ++ Al BSF region formed using a 1064 nm Nd:YAG laser with 100 µs pulse length and a dielectric stack of ALD  [43]. Secondary Ion Mass Spectroscopy (SIMS) results indicated a junction depth of 1-2 µm and a peak Al concentration of 5 × 10 19 cm −3 , which is well above the solid solubility limit shown in Fig. 4, another benefit of laser doping. Sheet resistances were reported to be in the range of 300-400 Ω/sq, significantly higher than expected if all Al atoms in the silicon were electrically active as calculated using EDNA2 [48]. Despite the high sheet resistance, efficiencies over 21% and > 82% fill factors have been achieved, indicative that the laser doped regions were effective as local back-surface-field (LBSF) regions [49].  Table I. It is shown that sheet resistance values of less than 20 Ω/sq are achievable using this dielectric source method.
One more effective option to utilize Al dopant atoms is by forming a deeper laser doped region through the use of a Q-CW or CW laser (as described in Section II). Using a 355 nm mode locked Q-CW laser, Cornagliotti et al. demonstrated a junction depth of 5 µm for an LBSF region formed by laser doping through a dielectric stack containing a 10 nm thick layer of ALD Al 2 O 3 as the dopant source with sheet resistances as low as 9 Ω/sq [55]. A key feature that was noted in that work was that the p ++ regions formed not only extended deep into the substrate but also laterally beyond the opening in the AlO x layer (see Fig. 5). This was not observed using an identical process on a wafer with SiN x passivation layer and boron SOD source. It was hypothesized that the lateral doping beneath the AlO x layer was due to thermal diffusion of the Al dopants into the silicon from the AlO x layer in the solid state. This feature could potentially be beneficial in shielding minority carriers from laser-induced defects that form at the edges of the laser openings.

V. LASER DOPED SEMICONDUCTOR FINGER SOLAR CELLS
Semiconductor finger solar cells (SFSC) were developed in 2006 as a means to implement a selective emitter structure with screen-printed contacts in production [56]. The SFSC made use of narrow heavily doped channels on the surface of the solar cell running perpendicular to the metal fingers providing a lateral conductive path between the fingers. This meant fingers could be placed further apart with a multitude of resulting benefits including less optical shading, self-aligned screen-print metallization, and reduced metal/Si interface (as the metal fingers only contacted the silicon where they intersected the SFs). Originally the SFs were formed in a similar manner to the buried contact solar cell (laser grooving followed by thermal diffusion), later, laser doping was implemented to form the SFs whereby an SOD was applied, and laser doping was performed through the SiN x dielectric layer [57], [58], making the technology more commercially viable and also compatible with lower quality wafer types. In the advanced laser-doped semiconductor finger solar cell concept, the SFs were plated with a thin layer of metal after the screen-printed fingers were formed, improving contact resistance at the intersecting points and improving the lateral conductivity of the SFs meaning the screen-printed fingers could be placed further apart again. A schematic of the advanced laser-doped semiconductor finger solar cell is shown in Fig. 6.

VI. LASER DOPED INTERDIGITATED BACK CONTACT CELLS
IBC cells offer several advantages over conventional cells with contacts on both front and rear, including reduced shading loss and series resistance, decoupling of the optical and electrical optimization, a uniform appearance, a simpler interconnection process, and a higher packing density leading to higher module efficiencies [59].
Despite the high efficiency potential of IBC technology, the complexity of the fabrication process has been a major factor in the limited commercial success of the structure that only accounts for less than 2% of the world's solar cell market share [10]. The low market share is predicted to continue for the next decade, remaining below 5% in the year 2031 [10], largely owing to the high production cost as well as the transition of the industry toward bifaciality for PERC and TOPCon cells instead, which offers a higher potential for cost per watt improvements [60]. The complexity associated with IBC cell fabrication lies in the requirement for regions of both positive and negative polarity to be formed in an interdigitated pattern on the rear of the device and for these regions to be contacted with metal whilst avoiding shunting between positive and negative contacts. The interdigitated pattern has typically been facilitated via high-temperature thermal diffusions either through barrier masks to form selectively diffused layers as in the passivated emitter rear floating junction cell [61], [62], or through the selective removal of a diffused layer after its formation, using methods such as laser ablation or photolithography and wet chemical etching [61], [63], [64], [65]. These methods also typically include the use of high temperature thermal oxidations to form oxide masks. As an example, the interdigitated buried back contact cell developed at UNSW required seven high-temperature diffusion or oxidation steps [66]. In 2004, SunPower commercialized the A300 rear point-contact solar cell concept developed by Stanford University [67] replacing photolithography with screen-printing technology to form the Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.  patterned diffusion masks, achieving production efficiencies of 21.5% [68]. SunPower now mass produces cells with efficiencies above 24.5% [69] and module efficiencies up to 22.8% [70], although these are typically considered a premium product due to the high lifetime substrates required and the processing that is still relatively complex compared to the more conventional PERC solar cell structure. The use of laser doping in IBC has the potential to significantly simplify the fabrication process of IBC cells further. Replacing high-temperature processing steps with laser doping also facilitates the use of wafer substrates that are sensitive to oxygen precipitation.
In recent years, several research groups have explored the use of laser doping in IBC cells. Most notably, the University of Stuttgart achieved 23.2% efficiency in 2016 where laser doping was used to form both n-type and p-type contacts on the rear, along with laser ablated contact openings to replace photolithography [71] (shown in Fig. 7). To form this cell structure, laser doping was carried out with a sputtered boron dopant source on the rear surface of an n-type wafer to create the p ++ emitter regions. A POCl 3 diffusion was performed to create a high sheet resistivity n-type diffused layer on both sides, the front surface diffusion becoming the front surface field and the rear surface diffusion, and its PSG layer being capitalized on to provide a dopant source for the n ++ laser doped base contact regions. Following PSG removal, a silicon etch process was performed to remove the phosphorus doped dead layer from each side and remove the possible shunt path between emitter and base. The efficiency of the cell was limited mainly by bulk resistance and bulk recombination, ahead of the losses associated with the laser-doped regions. Other notable examples where laser doping has been utilized in IBC cells are presented in Table II.
Another method of interest that has the potential of simplifying the IBC cell processing even further is the use of laser doping to overcompensate a diffused emitter layer. In this method, dopants of opposite polarity to the diffused surface are introduced via laser doping, in sufficient quantities to overcompensate the diffused surface region and form a contact to the underlying base region of the same polarity. This eliminates the need to selectively remove the diffused layer or perform masked diffusions and enables a high emitter coverage fraction to be maintained, as there is no gap between the p ++ and n ++ region. This process was demonstrated by UNSW using a 532 nm CW laser to introduce boron p ++ regions on phosphorus-diffused p-type wafers [73]. Experimental results indicated that there was minimal tunnel shunting between the heavily doped n ++ layer and p ++ LD region [74]. A proof-of-concept IBC cell was demonstrated using this method, whereby both polarities of contacts were formed via laser doping through a rear phosphorus emitter on a p-type substrate [75]. Only a single thermal diffusion step was required which simultaneously formed the rear n + emitter region and n + floating junction on the front side.

VII. TOPCON CELLS
The TOPCon cell structure is currently gaining market share due to the performance enhancements it offers over PERC. In this architecture, the rear surface of the n-type wafer consists of a thin tunnel oxide layer and n-type doped polysilicon layer contacted with silver to form a passivated contact with very low recombination losses. The front surface consists of a homogenously doped boron emitter passivated with a Al 2 O 3 /SiN x dielectric stack that acts as the anti-reflection layer. There is a significant discrepancy in the J 0 associated with front and rear surface, with the efficiency of the cell being limited by the front surface design. A selective emitter design would appear a sensible step to drive performance enhancements in TOPCon, with modelling suggesting an efficiency gain of 1% abs being possible [76].
So far selective boron emitters for TOPCon cells have only been demonstrated in a laboratory setting using chemical etch back approach [77]. In that study, to achieve the low sheet resistances of 65 Ω/sq in the contacted regions a long, energy intensive high-temperature boron diffusion was required. The implementation of the selective emitter resulted in a 0.2% abs efficiency gain compared to the control devices with a homogenous boron emitter.
The application of laser doping seems to be a logical next step to implementing an industrially relevant process for selective emitter formation on the front surface of a TOPCon cell. The presence of the Al 2 O 3 layer lends itself to the possibility of using the Al dopants present in the layer for the laser doping process as discussed in Section IV. However, it remains to be seen whether additional dopant sources would be required due to the limited solid solubility of aluminum in silicon. The potential for a post laser doping anneal at a more favorable temperature for the solid solubility of Al in silicon should be explored as an option that could be used to activate some of the inactive Al dopants and therefore achieve lower sheet resistances.

VIII. LASER-INDUCED DEFECTS AND PASSIVATION METHODS
Even though laser technology is a very attractive approach for a plethora of applications in silicon photovoltaic manufacture, its application requires a considerable amount of finetuning of laser parameters to reduce the detrimental effects that accompany laser processing. In particular, laser processing can adversely affect device performance by reducing minority carrier lifetime [47], [78], [79], [80], [81], [82], increasing leakage current [78], [83], [84] and diminishing carrier transport [81], all depending on the surface preparation, pre-existing dielectric stacks and the applied laser conditions (pulse duration and energy density). Even though first reports on the effects of laser processing on device performance go back over five decades [83], conclusive identification of some of the introduced point-defects causing much of the diminished device performance and so-called laser-induced defects, is still lacking in the literature. Diverse characterization techniques have been used to investigate the electrical, optical, and structural properties of such defects. Amongst these methods, deep level transient spectroscopy (DLTS) [85], [86], [87], photoconductance decay [78], [80], Hall effect [88], [89], photoluminescence [80], [90], [91], [92], [93], [94], scanning electron microscopy (SEM) [93], transmission electron microscopy [6], [47], [78], [95], [96], [97], SIMS [47], [80], [98] X-ray diffraction [99], and Raman spectroscopy [100], [101], [102] are the most widely applied. From an early stage, studies demonstrated that frequent melting and recrystallization processes caused by the pulsed lasing caused the formation of extended defects in the otherwise perfect lattice as well as the redistribution of nearby impurities. From DLTS   TABLE III  IDENTIFIED LASER DOPING INDUCED DEFECTS electrical characterization studies, trap states revealing point defects associated mostly with oxygen-vacancies, divacancies, and some high-order clusters have been reported [87], [100], [103], [104], [105], [106], [107], [108], [109], [110], [111]. However, conclusive identification of most of the formed defects is still lacking and only ambiguous identifications have been reported. In Table III, a summary of the point defects introduced on p-type and n-type silicon upon laser irradiation with their corresponding suggested chemical identities and energy levels is presented. Similarly, using microscopy techniques, extended defects such as dislocations, micro twins and swirl-like defects have also largely been identified [78], [83], [95], [96], [100]. In addition, some other defects in the form of voids within the bulk of the melted region have also been observed using SEM imaging for slow laser scan speeds when using a CW laser [34]. It was found that void concentration and size increase when low laser speeds are used, although their density was also dependent on the dielectric layers and the deposited dopant source.
In general, however, the nature of the commonly detected defects will greatly depend on two factors, 1) the initial concentration of extrinsic impurities (carbon, oxygen, nitrogen, dopant, etc.), and 2) the degree of structural imperfection of the silicon crystal given that this will directly impact the threshold melting of the material [86].
It is also clear from the literature that significant defect formation is only observed when laser irradiation surpasses the silicon melting threshold which can be regulated by the laser fluence, and that recrystallization velocity is a critical parameter for defect formation. In particular, dislocation density and oxygen introduction have been reported to have an impact on the surface electrical and optical properties of solar cells [94], [116]. Recently, Sun et al. [78] determined that less than 20% degradation of the surface can be attained when restricting the laser-induced dislocation concentration to ∼10 6 cm −2 . In that study, the degradation was analyzed in relation to surface carrier lifetime, surface drift mobility, and surface conductivity, while dislocation density was determined using a combination of isotropic etching and SEM imaging.
Several methods for the reduction of laser-induced defects have been reported in the literature. The initially explored and obvious alternative has been the optimization of laser processing conditions, which have been demonstrated to have a strong effect on defect formation [99], [117]. Thermal stress appears to be a significant factor in laser-induced defect formation and can be minimized by reducing the thermal cycling. An easy approach to this is through the use of modified Q-switched lasers or preferably CW or Q-CW lasers [118]. In this way, repeated melting and resolidifying of the silicon between pulses is minimized or avoided, and defect generation often observed along the edge of the laser-doped region can be suppressed [119]. A further benefit arising from the use of CW or Q-CW lasers is the deeper junction formation that is possible, enabling better shielding of minority carriers from the laser-induced defects generated at the surface.
Another source of thermal stress is the thermal expansion mismatch between the silicon and the overlying dielectric layer in the case where laser doping is performed after dielectric deposition [120]. Performing the laser doping prior to dielectric deposition (directly on the silicon surface) can avoid this issue, as discussed in Section III-A. An alternative approach is to modify the dielectric stack to reduce the thermal expansion mismatch through the incorporation of a thin SiO 2 layer between the silicon and SiN x [1], [45], [119], [121].
Many reports have focused on post-thermal treatments for defect annihilation and hydrogenation. This is because, many of the formed defects have been shown to disappear after thermal treatments in the temperature range 500-700°C, and to react with atomic hydrogen at temperatures between 100 and 300°C [86], [106]. Some examples of these include low H ion implantation [122], laser annealing [93], rapid thermal annealing [105], and advanced hydrogenation methods [123], [124], [125]. In practice nonetheless, where laser processing is used for doping and contact opening, defect reduction approaches heavily rely on the application of H-rich dielectric coatings either prior to or postlaser processing, which are often combined with a high-temperature "fast-firing" step depending on the device architecture. For example, the now dominant selective emitter PERC architecture in the industry uses the PSG layer as a dopant source prior to a PECVD SiN x :H deposition which is carried out at about 400°C in the presence of hydrogen-containing gases. This mild thermal process and the H-containing dielectric serve as H-plasma treatment and surface passivation of the laserinduced defects all within the same step of the manufacturing process. Following this, the fast-firing process carried out at temperatures between 700 and 900°C for contact formation after the screen-printing step further annihilates the laser-induced defects while injecting atomic hydrogen to passivate most of the remaining active defects.

IX. CONCLUSION
Laser-assisted doping of semiconductors has been studied for many decades, and its benefits have been evidenced abundantly in the literature. Progress in the development of faster, cheaper, and more reliable laser sources used in the semiconductor industry has enabled their worldwide adoption. In the silicon-based solar industry, however, it was only in the past half-decade that laser doping became the preferred technology for forming selective emitters on industrial solar cells. This adoption has been driven mainly by the adoption of higher quality silicon wafers resulting from price reductions and the implementation of rear-passivated concepts (PERC/PERT/PERL technologies), which place a higher weight on performance on the surface recombination currents.
Despite the availability of a number of approaches for doping using lasers, the industry-chosen method for selective emitter formation continues to be laser irradiation after the POCl 3 gas phase diffusion and prior to dielectric deposition. It may be possible that alternative dopant source methods (e.g., from dielectric layers) become more widely used in the future if plating metallization technologies do increase their market share in the coming years as predicted by the latest ITRPV report, or if laser doping becomes more commonly used to form BSF regions on PERC cells or a selective emitter on the front surface of TOPCon cells. In the case of IBC technology, some of the laser doping advantages have been reduced due to the continuous progress in the now very mature screen-printing technology (tools and pastes) capable of simultaneous doping, which still has the benefits of high throughput and simplicity. It may be that a combination of laser doping technology maturity leading to higher throughputs combined with the much-needed metal reduction consumption in the industry will lead to more widespread adoption.
There is still some work to do in the space of laser-induced defect identification as some of these are still unclear in the literature, however, several postprocessing approaches have already been adopted to reduce the detrimental impact of these defects via defect annihilation or hydrogen passivation leading to satisfactory results.