The Additive Manufacturing of Glass: A Critical Review

Abstract

This paper presents a critical review of additive manufacturing (AM) techniques applied to glass, elucidating the current state of the field and identifying key challenges and opportunities. The review identifies various AM methods, as applied to glass fabrication over broad length scales. Critical aspects such as material considerations, process parameters, and postprocessing techniques are presented, offering insights into the evolving landscape of glass AM. A particular emphasis is placed on semi-solid glass processing. The paper assesses the achievements and limitations of existing methodologies, paving the way for future advancements. This review serves as a resource for both researchers and practitioners in the emerging field of the additive manufacturing of glass, including applications in the domain of architecture and art.

1. Introduction

1.1. Glass and Its Relevance

Overall, the unique combination of properties offered by glass makes it an indispensable material in numerous industries and everyday applications. Glass has made numerous contributions historically [1] and continues to contribute significantly to modern life and technology. The principal production routes for glass are presently the float glass process, glassblowing, and flameworking techniques. The float glass process involves pouring molten glass onto a bath of molten tin, resulting in a uniform sheet. Glassblowing involves shaping molten glass by blowing air into it through a blowpipe, allowing intricate shapes and designs by either manual manipulation or formation with a mould. Flameworking operates in a similar manner, instead softening glass for manipulation with a flame. Whilst there are exceptions to these processes (i.e., the formation of glass bricks, or glass fibres), fabrication methods for glass have not been greatly disrupted in a manner similar to other industrial materials. The relatively recent emergence of the additive manufacturing (AM) of glass presents significant advances for glass as both a primary and secondary (i.e., reinforcing) material. The review herein is concerned with the use of glass as a primary material [2,3,4,5]. For example, an additive manufacturing process could potentially produce complex components exclusively from glass and potentially reveal a new paradigm in the way glass is utilised. Examples include complex shapes, load bearing structures, components that incorporate free space, and ‘mechanical metamaterials’ [6], with examples presented in Figure 1.

In addition, the AM of glass can also make a significant contribution to the reduction in carbon footprint, with recycling removing a reliance on mining or alteration in natural landscapes. In Australia alone, the volume of glass waste entering landfill is a significant issue, the scale of which continues to grow [7]; consequently, new processes that can utilise waste glass are critical. The prospect of waste glass as feedstock for additive manufacturing is a key facet of sustainable processing, whilst also allowing upcycling in applications for the quantity of waste glass being produced globally. The use of waste glass presents multiple benefits, including reduced energy consumption over virgin material [8] and a low-cost feedstock for commercially produced objects.

1.2. Glass as a Processable Material

As elaborated in the synopsis of techniques which are being currently explored for glass AM (in the following section), the processes for glass AM utilise the unique thermal behaviour of glass in order to produce an object. This is not coincidental, as the thermal properties of glass have been utilised in glass forming for centuries [9]. The unique relation between temperature and viscosity allows glasses to be shaped in a viscous liquid state. This is depicted in Figure 2, which presents viscosity (for soda–lime glass) as a function of temperature. The range of viscosity (given that the y-axis in Figure 2 is a logarithmic scale) for glass covers more than 12 orders of magnitude over a range of approximately 1000 °C.

A gradual change in viscosity according to temperature allows a suitable working temperature to be chosen, balancing various factors such as the precision of material deposition (influenced by viscoelastic response, die swell, etc.) and the speed of printing (influenced by the pressure required for adequate extrusion, the rate of viscosity variation in material, etc.). A temperature is chosen at a suitable trade-off, such that the glass is ‘malleable’ and easily deposited but will not deform under its own weight and lose complex details. While extremely variable between techniques, a temperature relating to the working point of glass ($~{10}^3 \mathrm{P}\mathrm{a}·\mathrm{s}$) is commonly chosen as a starting point, from which smaller qualitative adjustments are made to optimise material flow and stability after deposition [10]. For example, observations during an extrusion-based additive manufacturing process may identify that glass is not holding its shape after extrusion, failing to capture precise geometry. To remedy this, the temperature of extrusion may be slightly decreased, causing extruded glass to become more viscous and remain in position once deposited. The decrease in temperature must strike a balance, however, as continuing to increase glass viscosity will eventually introduce issues with the mechanism of extrusion.

Figure 3 reveals a smaller range of temperatures, and viscosity as a function of temperature, for various common glasses. It can be seen from Figure 3 that there is minor variation in soda–lime glasses (depending on colour, with colour arising from controlled minor/trace changes in glass chemistry), whilst so-called artist glass (such as Bullseye glass) has a lower viscosity across a range of temperatures, and borosilicate glass has a markedly higher viscosity across a range of temperatures. In contrast, quartz glass (SiO₂) is a type of glass that does not contain additives, resulting in working temperatures that fall outside the ranges depicted in Figure 3. The results presented in Figure 2 and Figure 3 may also be calculated using a convenient user tool developed by Korber [11]. The tool has a user-interactive web site that draws from the SciGlass database, providing the viscosity and physical property of glasses as a function of their composition.

The thermal properties of glass are leveraged in the direct AM of glass. A laser or heated nozzle is used to heat the glass feedstock to a suitably workable temperature, such that the material may be readily extruded or melted upon the build area. Deposited as a hot, viscous liquid, the glass may fuse with a previously deposited layer, necessary for the structural integrity of the finished object. As a fused deposit begins to cool, it also begins to undergo an increase in viscosity towards a solid form. This ensures structural integrity during AM, as layers can be stacked vertically without the earlier layers deforming due to the weight of subsequent layers. Whilst this process is elaborated further below with the relevant technical detail, it is mentioned here to draw attention to the many variables (including material variables) and processing variables (i.e., AM parameters) that must be optimised in order to ensure successful—and high-quality—glass AM.

While the temperature–viscosity relationship of glass enables direct AM processes, secondary thermal behaviours also require consideration. A commonly observed behaviour of glass materials is their tendency to fracture whilst experiencing a large or rapid temperature differential. This presents an issue with direct glass AM processes, as these methods deposit a ‘hot’ material upon a ‘cooled’ (lower temperature) prior layer. As direct AM processes continue, an appreciable thermal gradient may develop within the produced object, resulting in stresses from thermal expansion which may cause fracture or poor mechanical performance. There has been little consideration of this issue in the existing literature, with thermal modelling focused upon initial melting and material deposition [10,12,13], rather than gradients in the produced object. To date, a heated chamber enclosing the AM build area has typically been used in place of a detailed discussion of the thermal behaviour of AM-prepared glass components. In holding a produced component at an elevated temperature, a critical temperature differential is avoided, annealing the object during production [10,14,15].

1.3. General AM Overview and Examples

Still in relative infancy, the field of glass AM has seen limited reviews to date [2,3,4,5]. The contemporary glass AM techniques documented have—thus far—been analysed mostly for functional and high-resolution applications; i.e., gradient optical lenses, microfluidics, and microelectronics are regularly considered [16]. The high technical requirements of such functional applications create a requirement that highly values component resolution and optical clarity. Such functional applications are but one aspect of those relevant to glass, albeit they demonstrate new utilisation for glass through precise geometric fabrication. Similarly, at the larger end of manufacturing length scales, there is also a potential environmental and social impact of larger-scale (and lower-resolution) techniques, applied to art and architecture. At length scales of >>10⁻³ m, aesthetic and mechanical properties become the prevailing features used to examine glass AM technology. Artistic and architectural works must be readily producible at a suitable length scale, with adequate structural stability and integrity for use. Certain glass AM techniques may even enable new aesthetics to be utilised and incorporated into the artistic process; given that the human eye is much less discerning when it comes to optical clarity, the variation in some AM techniques is made more acceptable. In the present review, an analysis of glass AM at length scales that may be termed larger-scale (i.e., where mechanical properties are considerations) is also introduced, providing a new perspective on studies already existing in the field.

1.3.1. Relevance to Art and Design

It has been noted that recent research in the domain of glass art has included the creative development of digital manufacturing techniques [17]. This includes so-called vocabularies of form which have arisen from incorporating the following approaches (for example): (i) waterjet cutting into stained glass practice, (ii) computer-based image-making into 2D and 2.5D screen-printed and sandblasted imagery in kiln-formed glass; and (iii) 3D printing in cast glass to create previously unobtainable forms.

The application of glass AM is not yet widespread in artistic practice, and therefore, what have not been replaced—to date—include traditional analogue techniques. However, glass AM has the potential to remove the associated constraints presently required to digitally translate a derived form into a resulting object in glass. At present, to turn a 3D model into cast glass, a polymeric 3D print is usually created and realised through gravity-based lost-wax kiln casting—a process largely unchanged since its invention many centuries ago. Intermediary materials such as silicone, wax, plaster, and other refractories are used in a sequence of reversals between positive and negative, each process with a loss of fidelity to the original form [17]. In the glass art sector, a gap in knowledge exists, and there is a demand for makers who can realise digitally derived designs and machine-fabricated forming processes. Benefits of the direct production of 3D forms via the additive manufacturing of glass include the following:

• Rapid prototyping, allowing the translation of an idea into a form within a short period of time (i.e., hours);
• The ability to realise shapes and structures, from direct manufacturing in 3D, that are otherwise unobtainable from conventional glass processing methods;
• The ability to realise detail in ‘internal’ structures that is not possible through methods analogous to the lost-wax approach;
• The ready combination of colour transitions, and diverse features.

Some examples are provided in Figure 4 and Figure 5.

1.3.2. Relevance to Architecture

The emergence of AM in general has already received and made considerable attention and impact in the domain of architecture and construction. There are numerous reports regarding the AM of cementitious materials and concrete [18], and even a 3D-printed bridge in Amsterdam fabricated from stainless steel [19]. The application of AM in architecture and construction has predominately focused on the following: (i) demonstrating the technology, (ii) showcasing flexibility in remote construction, and (iii) demonstrating the speed of automated construction. However, the emergence of large-scale glass AM can open new architectural possibilities and catalyse innovation areas that may include sustainable façade design. For example, the potential of using recycled glass in the construction of buildings can be explored through additive manufacturing. Despite the development of large-scale AM processes for construction with the aforementioned materials [20,21], the development of large-scale 3D-printed glass has not been widely explored to date. The potential to apply recycled glass at a large scale to building façades therefore provides possibilities for the circular economy and sustainable building design.

Architectural tectonics (the integration of skin, structure, jointing, and material) has evolved in response to the available processes of construction. With an innovative new process for fabricating glass, there is significant potential to explore new architectural and structural design possibilities, with the early stages of this exploration shown in Figure 6. This will reduce the number of parts and trades, improve building performance, and generate new design potential. Three-dimensional printing has the capacity to create complex forms, which enable efficiencies of material through structural and design optimisation and levels of detail which are not feasible in standardised construction.

2. Glass AM Techniques

In an examination of existing techniques of glass AM, a distinction is drawn between two major methods of processing. These are the direct and binder-based processing methods. A summary of these methods is presented in Table 1. Unless specified, ‘2.5D’ processing is used in the currently explored methods. These create a 3D object by separating its geometry into discrete horizontal slices, corresponding to a layer height. Each of these 2D slices is stacked and deposited vertically, creating an object layer by layer. Such a technique greatly simplifies the programming of the deposition procedure, at the cost of some more complex geometry. Mainly, this process creates some difficulty in producing the undersides of a slope (i.e., overhangs) and features which extend unsupported by previous layers of material (i.e., bridging). Such features may still be possible with 2.5D processing but likely come at the detriment of surface finish or material wastage as a support structure is required. The alternative to this is true non-planar printing, where deposition is allowed to move beyond a horizontal plane. The programming of this is non-trivial and largely unexplored in the reviewed techniques.

2.1. Direct Methods

Direct methods involve the deposition of glass material upon the build area, resulting in a solid glass model. Objects produced with this process are in a complete, solid glass form (i.e., net shape or very-near-net shape) directly after the AM process finishes.

2.1.1. Directed Energy Deposition (DED)

Directed energy deposition (DED) utilises a focused energy source to fuse a separately fed feedstock material to a constructed object. Such a method has been explored for glass AM, mainly using a laser for fusing a feedstock material of solid glass filament. Depending on the level of detail required, filaments may be drawn by hand or by some mechanically assisted process such as those required for lab glassware or optic fibres. CO₂ lasers are typically used, as they generate the wavelength of radiation that is most conducive to the melting of glass, whereby the glass feed material is opaque and has minimal reflections [22,23]. The ability to melt glass using a laser, along with the associated efficiency, is an area that requires focused future work. In recent reports, preliminary investigations have also been undertaken, exploring the use of a flame for the fusing (melting) step [24,25].

Depicted in Figure 7a, the DED process proceeds by focusing the energy source to generate a pool of molten glass, to which additional feedstock can be added. Via computer numerical control, the melt pool is scanned across a user-defined path with the continuous addition of feedstock, depositing a pattern of fused glass. The currently explored DED techniques are mainly ‘2.5D’ in nature, where the heat involved in deposition fuses single layers of material to previous layers. Some non-planar deposition techniques have been explored [26,27] but are still in their infancy at the time of review.

With the appropriate process optimisation, DED is capable of excellent fusion between layers, with transparencies nearing those of cast glass [29,30,31,32]; however, a thorough treatise regarding the mechanical properties of these objects is yet to be written. Variables for process optimisation typically include laser power (W), the material feed rate (x/min), and laser scan speed (mm/s). These process parameters are varied so that the melting of glass feedstock is balanced in terms of sufficient material fusion to avoid the overheating of the molten region. Overheating may cause the vaporisation of the glass material, creating bubbles within the printed object [12,29,30,31,33,34,35]. Further process adjustments, such as focusing the laser beneath rather than above the feedstock, are also being explored to improve melting, with a reduction in material vaporisation [13]. The glasses explored to date using the DED method include those elaborated in Table 1.

The excellent fusion in the DED process comes with a trade-off for the precision and resolution of printed features. In the currently explored processes, interactions between the melt pool and feedstock are seen to create inconsistent deposition and deformation of existing glass material. As illustrated in Figure 7b, presently explored DED-manufactured objects exhibit a fluctuating outer surface with a tendency of the liquid melt pool to smooth over the finer details of the object, with such characteristics also being a detriment to overhanging features. In 2.5D processing, overhangs require a stepover of material, where deposition occurs partially unsupported by the previous layer, relying on the quick cooling and high viscosity of material to remain in place. These unsupported steps will likely see a similar smoothing and sagging of fine detail, limiting the possible overhang angle.

The possibility exists for improvements to this process, such as enhanced object resolution with tighter control of the melt pool [26,34,36]. This is at odds with the feasible scale of the process, as a smaller melt pool limits the rate of material deposition. Print times become infeasible for larger objects, as the DED machine deposits less material for each unit of movement. Solid 3D objects would require additional scanning of the melt pool upon each layer, causing individual layer times to increase rapidly.

Additionally, DED objects produced at a larger scale present a need for further process design. Unlike metal DED (which may also involve a spooled or blown powder feedstock), the glass feedstock may have limitations in the associated feeding mechanism. Exceeding a critical diameter, glass filaments become inflexible (rigid, yet brittle) and cannot be easily wound around a spool, for example. Considering high deposition rates, a trade-off must be made in the feedstock used. Large inflexible filaments facilitate higher deposition rates, but there is issue when objects exceed the material provided in one length of filament. There must be consideration of how new feedstock material is added while not interrupting the process of material deposition. Alternatively, limiting the filament diameter allows spooling and the easier delivery of the feedstock material [26,34,36,37,38] but must facilitate higher feed rates to achieve a material deposition rate suitable for large-scale objects. Powder-fed DED may also mitigate this issue but has seen limited exploration in glass AM at the time of review. Such a process has been demonstrated only for single-layer glass structures, without the analysis of a multilayered 3D object [39,40].

2.1.2. Selective Laser Melting

Selective laser melting (SLM) techniques direct an energy source across a bed of powdered feed material. An early version of this approach was attempted in art by Markus Kayser, whereby focused light from the sun was used to melt desert sand [41]. Later processes followed the more developed procedures for SLM, using a laser as the energy source and a glass powder feedstock. An object is created by scanning the energy source across the powder bed to fuse the feedstock in a desired pattern, as depicted in Figure 8. Vertical features are obtained by lowering the object in the bed and redistributing a thin layer of powder feedstock on top. Therefore, 3D features are built layer by layer, fusing the newly distributed feedstock upon the existing layers. The redistribution of the powder bed presents a unique advantage of SLM methods, as any unfused powder may support the model as it is built. Overhanging and bridging features may be generated for ‘free’, where other methods would need support structures for this geometry.

Present SLM methods demonstrate the ability to produce solid glass objects; however, these methods introduce multiple issues for large-scale printing. The processes explored to date rely heavily on feedstock materials of controlled size, shape, and composition. SLM methods have typically used high-grade glass powders with a spherical particle shape (for better flowability and redistribution) and size ranging from ~1 to 500 µm [22,42,43,44]. At the present stage of research, methods are not being explored for a low-cost feedstock such as recycled cullet. There are likely to be issues at scale due to the sourcing of high-grade feedstock, or the processing of low-cost cullet to such a form suitable for an SLM method. Definitive research is needed in this area.

Also, the powder bed results in the incomplete melting of the feed material and porosity in the molten glass. This causes undesirable traits in the finished object. Illustrated in Figure 9, this process creates a mostly fused core surrounded by a rough, opaque outer surface of unmelted glass powder [22,42,43,45,46].

This outer layer, as depicted in Figure 9, has poor transparency, causing printed objects to lack the typical aesthetic qualities desired in a glass object. Based on research presented to date in the field, it appears that a postprocessing procedure such as mechanical or flame polishing is needed for SLM-produced objects with specific aesthetic or optical requirements (although such postprocessing methods are not yet widely explored). Such surface modification/polishing procedures ultimately limit the complexity of models, as they then require adequate access to all surfaces. The poor fusion in this process also raises the need to better understand the mechanical stability of glass objects prepared by SLM. Defects within the solid object create points for stress concentration, increasing the likelihood of fracture. The flexural testing of SLM-produced samples has shown strengths significantly lower than those of float glass of a similar material [42]. This presents issues with objects of significant scale, as the weight of the object itself may exceed the inherent strength of the object possible with this technique. This merits a brief comment that metal–glass hybrids, where the proportion of glass powder was no more than ~30% in a glass–metal powder blend, have been successfully produced using SLM [47]. In the work, the resulting objects following fusion were metallic, with glass metallurgically fusing with the (majority) metallic feedstock.

2.1.3. Semi-Solid Processing

One may notice that the thermoplastic polymers (PLA, ASA, PETG) used in extrusion-based additive manufacturing (EAM) exhibit similar semi-solid thermal behaviour to that of glass materials. As depicted in Figure 10, upon extrusion out of a heated nozzle, both glasses and thermoplastic polymers commence in a ‘liquid form’ (which is, in the case of glasses, defined as a low enough viscosity to allow flow characteristics) and increase in viscosity until the material is ‘semi-solid’, fixed in the so-called net shape position. Such behaviour is combined with computer numerical control, allowing the creation of complex 3D objects with the fusion of one deposited layer to the next, and changes in object geometry over the course of multiple layers (gradients and overhangs, bridging, etc.).

Correspondingly, knowledge of the extensive developments in the field of polymer AM, including advancements in hardware and particularly software, may be leveraged to create objects from glass. The most direct of these approaches follows polymer EAM. Such processes involve hardware related to polymer EAM methods, insofar as the process facets for glass EAM are synonymous with those of polymer EAM, modified for silicate materials. For glass EAM, the process needs to facilitate high-temperature processing, permitting the extrusion of silicate glasses. Consequently, these processes must include refractory components and kinematics designed to withstand the temperatures at which glass has low viscosity. Such processes are the focus of this review and are discussed in more detail in Section 3.

2.1.4. Low-Temperature Semi-Solid Processing

An alternative approach to modifying the printing technology is to focus on adapting the material rather than the process. Therefore, glass EAM has also been explored for low-temperature glasses such as chalcogenide, phosphate, and heavy metal oxide glasses. These are glasses characterised by a relatively low glass transition temperature (e.g., ~188 °C for As₄₀S₆₀ chalcogenide glass [49]) compared to silicate glasses. Three-dimensional objects made of chalcogenide and phosphate glass have been produced with only a slight modification of commercial polymer EAM machines and software—such hardware is depicted in Figure 11. Where a spool of polymer feedstock would be located, filaments of chalcogenide or phosphate glass are fed into an extruder head. Through the use of common slicing software and the appropriate choice of extrusion temperature, objects are created using a typical workflow that is routine for polymers. Objects made of heavy metal oxide glasses have required more process design, using a custom heated and pressurised crucible to facilitate material deposition [50].

While it is desirable to take advantage of well-developed polymer printing and lower printing temperatures, reliance on exotic materials severely limits utility compared to silicate glasses. The explored low-temperature glasses have had limited application outside of optics and semiconductors. For social impact, such as in art and architecture, glass AM must find use outside of an engineering environment. Therefore, material availability and safety are concerns. There is not a significant waste stream to be utilised as with silicate glasses. Also, in particular, chalcogenide glasses have been observed to produce hazardous fumes while printing, requiring a fume hood enclosing the printing area [49]. Examples of models produced by EAM with chalcogenide glass are presented in Figure 12.

2.2. Binder-Based Additive Manufacturing

Binder-based glass AM is enabled by leveraging existing material fabrication techniques, by introducing glass feedstock into more readily ‘printable’ materials, such as photopolymers and colloid inks. The AM methods applied in this context vary rather greatly and are summarised in Table 2. Generated models form a green object, similar to those in ceramic manufacturing. Subsequent treatment processes remove the binder from the model and join the remaining feedstock into a solid glass object. Notable in this subset of processes is the significant shrinkage of objects after treatment, due to the complete removal of binding material, and the near-complete consolidation of the glass feedstock. The most explored binder-based additive manufacturing methods to date have been direct ink writing and light-based methods. Direct ink writing (Figure 13a) notionally involves the use of fine ground glass powder mixed into a slurry that may hold its shape after deposition. Computer numerical control guides a syringe which extrudes the glass slurry into complex shapes, with the resulting object ultimately sintered (at high temperature) to produce a glass solid [51]. The light-based methods (Figure 13b) use a glass feedstock or precursor in combination with a photopolymer binder. Controlled exposure to light (such as through existing stereolithography or two-photon polymerisation methods) cures the photopolymer in a desired shape, ready for high-temperature processing, where the binder is removed and the remaining glass is sintered into a solid object [52]. Limited research has been conducted on a similar EAM method, where a thermoplastic polymer binder is combined with glass powder [53].

Binder-based glass AM techniques can yield objects with excellent transparency and fine detail. The development of the glass + binder material and subsequent treatment processes has enabled completely solid glass objects, without residual organic materials or porosity [54]. The significant shrinkage of the printed object also enables higher-resolution features than typically possible with other material on the same machines.

Limitations in model size are expected with these processes due to the cost and complexity of the feedstock material and the need for the treatment of the green models. Exceeding a certain size, treatment procedures through temperature or chemical means become rather cost-prohibitive or may introduce problems such as uneven shrinkage or variance from strict tolerance. There have been investigations into improving these treatment processes [55,56], lowering the time in which and temperature at which the consolidation process takes place, but this research still only explores the smaller scales available in glass AM (~10 mm and below). Initial research has also been conducted regarding a hybrid technique, which may lessen the limitations of size in this process [57]. This involves a direct ink writing method depositing a layer of green glass feedstock. This method differs by following deposition with a laser melting process (as the energy input) that fuses the deposited glass in situ. Such a development may be applied to eliminate post-treatment procedures, though it should be noted that further research is needed to characterise the mechanical properties of objects prepared by such hybrid material approaches.

2.3. Synopsis of Reported Glass AM Methods

The following three tables (Table 1, Table 2 and Table 3) provide insight into the glass AM methods presently reported in the literature and relevant patents, along with their key characteristics in synopsis.

Table 1. Glass AM methods that employ glass feedstock materials.

Method	Feed Material	Deposition Width Scale (mm)	Length Scale (mm)	Advantages	Limitations	Related Literature
Directed Energy Deposition	Fused Quartz Filament	~0.1–4	~1–200	Excellent material fusion Excellent optical clarity	Lacks fine detail due to melting behaviour Feed mechanism needs consideration for uninterrupted printing	[28,31,32,34,36,37,38,40,58,59]
	Soda–Lime Filament/Powder					[14,22,26,27,29,30,33,39]
	Borosilicate Filament					[12,13,14,27,35]
	Germanate Filament					[60]
Selective Laser Melting	Fused Quartz Powder	~0.1–1	~1–100	Powder bed acts as support material, enables overhanging structures for ‘free’	Poor material fusion Reliant on high-grade materials Poor aesthetics/optical clarity without postprocessing	[61]
	Soda–Lime Powder					[22,42,43,44,45,61,62]
	Borosilicate Powder					[44,46,63]
Fused Deposition Modelling	Soda–Lime Filament/Cullet	~0.1–20	~1–~300	Highly cost effective Ability to produce complex structures	Surface finish (and opacity) dependent on deposition layer height	[10,15,64]
	Low-Temp Glass Filament/Pellets	~0.4 (Polymer printing)	~1–200 (Polymer printing)	Well-developed polymer AM workflow may be used	Severely limited utility of exotic glasses	[49,50,65,66,67,68]
Binder-Based	Various (See Table 2)			Excellent optical clarity Excellent fine detail Adopts workflow of well-developed techniques	Treatment processes required after initial printing before glass object is formed (debinding, sintering, etc.) Component shrinkage during treatment limits scale Expensive feedstock and preparation	(See Table 2)

Table 1. Glass AM methods that employ glass feedstock materials.

Method	Feed Material	Deposition Width Scale (mm)	Length Scale (mm)	Advantages	Limitations	Related Literature
Directed Energy Deposition	Fused Quartz Filament	~0.1–4	~1–200	Excellent material fusion Excellent optical clarity	Lacks fine detail due to melting behaviour Feed mechanism needs consideration for uninterrupted printing	[28,31,32,34,36,37,38,40,58,59]
	Soda–Lime Filament/Powder					[14,22,26,27,29,30,33,39]
	Borosilicate Filament					[12,13,14,27,35]
	Germanate Filament					[60]
Selective Laser Melting	Fused Quartz Powder	~0.1–1	~1–100	Powder bed acts as support material, enables overhanging structures for ‘free’	Poor material fusion Reliant on high-grade materials Poor aesthetics/optical clarity without postprocessing	[61]
	Soda–Lime Powder					[22,42,43,44,45,61,62]
	Borosilicate Powder					[44,46,63]
Fused Deposition Modelling	Soda–Lime Filament/Cullet	~0.1–20	~1–~300	Highly cost effective Ability to produce complex structures	Surface finish (and opacity) dependent on deposition layer height	[10,15,64]
	Low-Temp Glass Filament/Pellets	~0.4 (Polymer printing)	~1–200 (Polymer printing)	Well-developed polymer AM workflow may be used	Severely limited utility of exotic glasses	[49,50,65,66,67,68]
Binder-Based	Various (See Table 2)			Excellent optical clarity Excellent fine detail Adopts workflow of well-developed techniques	Treatment processes required after initial printing before glass object is formed (debinding, sintering, etc.) Component shrinkage during treatment limits scale Expensive feedstock and preparation	(See Table 2)

Table 2. Synopsis of binder-based glass AM methods.

Binder-Based AM Method	Feedstock Material	Deposition Width Scale (mm)	Length Scale (mm)	Related Literature
Fused Deposition Modelling	Silica particles + Thermoplastic polymer binder	~0.4 (Polymer printing)	~1–200 (Polymer printing)	[53]
Direct Ink Writing	Colloid of silica-based glass feedstock, rheological agents + Solvent	~0.1–1	~1–200	[51,54,57,69,70,71]
Light-Based (Stereolithography, Two-Photon Polymerisation, Direct Laser Writing, etc.)	Silica particles + Photopolymer binder	~0.0001–0.1	~1–200	[52,55,56,72,73,74,75,76,77,78,79,80,81,82,83,84]

Table 2. Synopsis of binder-based glass AM methods.

Binder-Based AM Method	Feedstock Material	Deposition Width Scale (mm)	Length Scale (mm)	Related Literature
Fused Deposition Modelling	Silica particles + Thermoplastic polymer binder	~0.4 (Polymer printing)	~1–200 (Polymer printing)	[53]
Direct Ink Writing	Colloid of silica-based glass feedstock, rheological agents + Solvent	~0.1–1	~1–200	[51,54,57,69,70,71]
Light-Based (Stereolithography, Two-Photon Polymerisation, Direct Laser Writing, etc.)	Silica particles + Photopolymer binder	~0.0001–0.1	~1–200	[52,55,56,72,73,74,75,76,77,78,79,80,81,82,83,84]

Table 3. A snapshot of patents related to glass AM, confined to those that employ solid-state glass processing (i.e., glass in the semi-solid/viscous state) methods.

Patent Name	Date First Filed	Assignee	Patent Number	Ref.
Methods and apparatus for additive manufacturing of glass	2015	Massachusetts Institute of Technology	US10266442B2	[85]
A kind of 3D printing equipment for chalcogenide glass element	2016	China Building Materials Academy	CN106116120A	[86]
Additive manufacturing systems and method for making glass articles	2017	Corning Incorporated	US20210101818A1	[87]
3D printing system for printing high melting temperature materials	2018	Micron 3Dp Ltd., D Swarovski KG (Location depending)	WO2018163006A1	[88]
Three-dimensional printing system	2020	Maple Glass Printing Pty. Ltd.	AU2021289614A1	[89]

Table 3. A snapshot of patents related to glass AM, confined to those that employ solid-state glass processing (i.e., glass in the semi-solid/viscous state) methods.

Patent Name	Date First Filed	Assignee	Patent Number	Ref.
Methods and apparatus for additive manufacturing of glass	2015	Massachusetts Institute of Technology	US10266442B2	[85]
A kind of 3D printing equipment for chalcogenide glass element	2016	China Building Materials Academy	CN106116120A	[86]
Additive manufacturing systems and method for making glass articles	2017	Corning Incorporated	US20210101818A1	[87]
3D printing system for printing high melting temperature materials	2018	Micron 3Dp Ltd., D Swarovski KG (Location depending)	WO2018163006A1	[88]
Three-dimensional printing system	2020	Maple Glass Printing Pty. Ltd.	AU2021289614A1	[89]

3. A Focus on Semi-Solid Processing for Glass AM

3.1. Process Overview

The descriptor of ‘semi-solid processing’ in glass AM encompasses methods that manipulate glass materials within a specific temperature range, typically between the glass transition temperature (T_g) and the melting temperature (T_m)—which, in the case of glass, are associated with changes in viscosity (over orders of magnitude) that permit glass processability [90]. This temperature range enables glass material to flow through a nozzle for deposition, and then quickly fix in place upon cooling. Semi-solid processing methods can enable the production of intricate digital designs using a variety of glass compositions—including recycled materials—through processes that include EAM.

Digital designs for semi-solid glass AM are typically defined using text files, known as ‘G-code’ files. With ‘G’ representing geometry, the G-code comprises a series of commands that translate to physical actions of the AM system. These commands autonomously manipulate and deposit glass to form desired geometries. Simple or complex G-code can be created directly using software including Rhinoceros 3D^TM, FullControl G-code^TM, and many other commercial and open-source software tools, offering intricate control over the manufacturing toolpath. Alternatively, 3D models (e.g., .OBJ, .STL, .STEP files) may be converted to G-code through what is known as slicing software (e.g., Simplify3D^® and PrusaSlicer^®). Slicing procedures facilitate the incorporation of geometric features such as infill density, infill pattern, multiple walls, and many other common AM parameters. Depending on the specific system employed, pausing and resuming printing may be enabled to print multiple parts and multi-contoured designs. Moreover, some semi-solid glass AM procedures support multi-coloured printing capabilities.

Following developed EAM techniques, the semi-solid processing of glass begins with the deposition of glass onto a printing platform. The printing platform acts as the first point of contact for the printed glass, preventing the displacement of deposited glass during production. In most cases, the first layer of glass adheres to the printing platform. This may be achieved by heating the printing environment and reducing the first-layer printing speed (as is common in polymer EAM). Compressing the initial glass layer onto the platform will also encourage bonding. Printing platforms are commonly chosen to be flat, high-temperature/refractory materials, capable of withstanding direct contact with molten glass. Alternatively, sheet glass may also be utilised as a platform, if the detachment of the component is not required. A high-quality first layer must exhibit both precise dimensional accuracy and strong adhesion to the printing platform. This is crucial, especially for larger 3D-printed components, as each build layer establishes the foundation for the subsequent layers.

Upon the completion of the first layer in semi-solid glass AM, the objective of subsequent AM fabrication layers is altered, since semi-solid glass is subsequently deposited onto existing glass. This allows for accelerated deposition speeds, further facilitated by the reduced energy requirements for interlayer bonding. Glass processing necessitates significantly higher temperatures than polymer EAM to achieve comparable (or even lower) viscosities. For soda–lime glass, typical settings involve nozzle temperatures of ~950 °C, with surrounding heated chambers of ~500 °C. Electronics are not typically able to withstand such a high-temperature environment; thus, semi-solid glass AM cannot utilise the layer cooling fans common in polymer EAM. Such fans, conventionally positioned near the nozzle outlet, play a critical role in rapidly cooling the material once it has bonded to the preceding layer, ensuring structural integrity and geometric precision. The rapid cooling of freshly deposited material is important in unsupported regions, such as bridged and overhanging sections (in the case of the printing of polymeric materials). To address this challenge in semi-solid glass AM, printing speeds may be limited as necessary to allow adequate time for the cooling of deposited glass layers. To obtain strong interlayer adhesion, nozzle temperature, ambient temperature, and printing speed are important variables to consider. High-quality interlayer adhesion is characterised by a homogenous bond, where layers are seamlessly fused together. Elevated processing temperatures provide a stronger bond through increased molecular diffusion and improved surface wetting between layers. Combined with slower printing speeds, this allows more time for the layers to merge and form a continuous bond.

The final step for the semi-solid processing of glass generally involves removing the deposited glass from the printing platform. As the glass cools and thermally contracts at a different rate compared to the printing platform, there is separation between the two. In cases where separation does not occur, a release agent such as ‘kiln wash’ may be applied before printing to ensure reliable removal, or mechanical removal may be employed.

Process control across the various methods [85,86,87,88,89] remains comparable, with differences primarily observed in hardware and feedstock type. There are two distinct methods of glass AM, depicted in Figure 14. The first involves a bulk high-temperature, low-viscosity feedstock (i.e., container of molten glass) with a secondary system controlling deposition. Conversely, the other method uses a low-temperature, high-viscosity feedstock (i.e., solid glass rods/filament), subsequently heated and deposited as needed.

3.2. Low-Viscosity Glass Feedstock Methods

Depicted in Figure 15, one of the earliest approaches to semi-solid glass AM involves a molten glass feedstock [10,15,64]. This method employs a crucible kiln mounted at the top of the system, storing the molten feedstock while printing. The filling (and replenishment) of the crucible kiln can be accomplished through two methods: the manual ladling of molten glass from a separate furnace or the gradual heating of glass cullet within the crucible. A smaller secondary kiln, named the nozzle kiln, is positioned at the base of the crucible kiln. Housed in the nozzle kiln is a heated ceramic nozzle, regulating glass flow via temperature. Unlike traditional EAM, which employs mechanically assisted extrusion, this method utilises gravity-fed processing. The nozzle extends into an annealing chamber, where a ceramic printing platform actuates in all dimensions, X, Y, and Z. During the printing of ‘System 96’ soda–lime glass, temperatures for the crucible kiln, nozzle kiln, and annealing chamber were set to 1040 °C, 1010 °C, and 480 °C, respectively.

During the initial heat-up process, the nozzle kiln is typically set to a lower temperature (~800 °C) to prevent unwanted flow. To begin printing, the nozzle kiln is set to 1010 °C, where glass will begin flowing spontaneously. As described in the general synopsis of semi-solid glass AM, the first-layer speed is reduced to promote adhesion to the platform. A layer height of 4.5 mm has been reported, controlled via the distance between the nozzle tip and the printing platform. Due to the gravity feed, the flow rate into the nozzle cannot be set directly; it is a function of the volume and viscosity of molten glass within the crucible. Therefore, layer width is dictated by printing speed and nozzle outlet size. A 10 mm nozzle was reported; however, the process may be varied to result in a line width of 19.5 +/− 3.5 mm and 9.5 +/− 0.5 mm. This may be achieved with printing speeds of 4.5 mm/s and 6.1 mm/s, respectively. Once the print process is complete, the nozzle must be cooled to <800 °C to stop glass flow. The process can produce components with exceptional aesthetic appeal, as is demonstrated in Figure 16.

3.3. High-Viscosity Glass Feedstock Methods

Solid glass rods (i.e., filament) or fibres also serve as a viable feedstock for the semi-solid processing of glass. With the use of a heating system (called a hot end), glass filament may be heated and extruded through a nozzle. Induction heating [88], joule heating [89], and direct flame methods [93] have all been used in hot end systems. These methods allow a filament to reach the temperatures and low viscosities required to achieve semi-solid processing.

In the case of using glass fibre, Pender [93] uses a flame to heat the fibre as it exits the nozzle. This allows the glass to be rapidly heated to over 1000 °C. Pender explored two methods of actuation using a robotic arm. One employs the arm to actuate the nozzle in all directions. The other uses the arm to actuate and roll the printing platform, while glass flows from a vitrigraph kiln [94] mounted above. Both methods are suited for achieving a unique drizzle effect with thin beads of glass, as depicted in Figure 17. The nature of the process does not enable detailed geometric accuracy, as the molten fibre is presented at some height above the built object. The fibre partially cools during deposition, causing it to buckle and coil as it reaches the object.

Moreover, the use of glass filament enables the use of retraction and extrusion movements, allowing material deposition to start and stop with a rapid response time. This approach is typified by the procedures of Micron3DP and Swarovski [88], and Maple Glass Printing [89].

There is little information available regarding the process employed by Micron3DP. The examples of glass models produced with the Micron3DP process showcase semi-solid processing’s ability to achieve a part resolution of ~0.1 mm (Figure 18). The process begins by extruding glass filament through an induction-heated nozzle, while the print platform actuates in all three directions within a heated chamber. Borosilicate and soda–lime glass have been processed. A similar example of well-resolved but undocumented glass AM using semi-solid processing is provided in Figure 19, revealing unique designs produced (at scale and in volume) by Swarovski for the Venice Exhibition in 2022.

The Maple Glass Printing (MGP) process employs a comparable EAM technique to the Micron3DP process, in which glass filament is extruded through a nozzle as the feedstock material. In the case of the MGP process, the hot end utilises a joule heater to reach temperatures suitable for semi-solid glass processing (to ~1150 °C). The method of actuation employs a platform that moves in a Cartesian coordinate system, within a heated chamber, with a standard build volume of ~170 × 200 × 300 mm (as seen in Figure 20a). For soda–lime glass materials such Bullseye^® glass (a piece prepared from such glass is shown in Figure 20b) or bottle glass, printing temperatures for the nozzle are between 900 and 1000 °C, enabling a typical flow rate of 21 mm³/s. The heated chamber allows the ability to anneal glass, the simultaneous production and annealing of the printed models, or programmed annealing (if the application or workpiece requires it) following the production of a model. Many glass materials have been explored by the MGP process, including soda–lime glass, waste container glass (i.e., recycled cullet), crystal lead glass, glassblowing waste (i.e., soda–lime ‘studio glass’ of various colours), and borosilicate.

The MGP process has yielded objects with a range of object parameters. Notably, there is a wide range of layer height selections between 0.1 and 5 mm (an example of high-resolution glass made using EAM is shown in Figure 21), and line widths ranging from 1.5 to 12 mm. An interchangeable nozzle allows the wide adjustment of these parameters, as the nozzle orifice diameter may be adjusted to better facilitate the desired printing scale (e.g., enlarging the nozzle orifice for greater layer heights and line widths or shrinking the orifice for decreased layer heights and line widths). Accompanying the wide adjustment of nozzle size, parameter change is also facilitated by changes in slicing software, altering parameters such as process speed and extrusion per unit of movement by the printer. The direct application of G-code may also be employed, which offers detailed control of the tool path (and was utilised in the production of the piece in Figure 20b).

3.3.1. Sustainability in Glass AM Through EAM and Example Models

When compared to the range of possible glass AM techniques (Table 1), semi-solid processing emerges as exceptionally versatile regarding feedstock material demands (though it should be noted that the finest length scales may not be accessible to the EAM approach). Additionally, waste glass also stands out as an important material choice (where virgin material is not essential) due to sustainability benefits. In many instances, the utilisation of 100% waste glass is feasible for EAM, without necessitating the addition of any virgin material. Examples of 3D-printed designs fabricated from fully recycled consumer wine bottle glass waste are depicted in Figure 22.

An example of an EAM-prepared glass model is shown in Figure 23. This is provided to highlight the breadth of geometries, contexts, and forms in which glass AM is being explored, including architectural models.

Moreover, glass AM through semi-solid processing has some additional benefits in terms of volume production and scaling, through what are relatively low energy requirements to power semi-solid processing. This efficiency is primarily attributable to high-temperature processing confined only to the nozzle. Glass processing methods at a production scale typically rely on large furnaces that are continuously holding the entire glass material at ~1200–1600 °C over multiple years [8]. Regarding the high-viscosity glass feedstock methods (detailed in Section 3.3), in such cases, the majority of glass feedstock is at room temperature without an energy cost; any energy usage is specifically for the facilitation of heating and deposition precisely where required on a produced object. There is secondary energy usage from the heated chamber and movement system, but these are still less energy-intensive than maintaining a furnace of molten glass.

In the context of the sustainability and recycling of glass, across all glass manufacturing techniques, devitrification is a concern when reusing glass. Devitrification, characterised by the transition of glass from its amorphous structure to a (partly or fully) crystalline structure, typically occurs due to the presence of impurities and exposure to specific temperature ranges conducive to nucleation and crystal growth [90]. In the case of the reuse of packaging glass, whilst it is known that such glass can be recycled many times (i.e., many dozens of times), there are often instances where, following a critical number of recycling cycles, glass may be more susceptible to devitrification. This phenomenon is, however, not well quantified in the field of glass AM, and most certainly is an avenue that merits future work. To date, issues of devitrification have not been critical concerns in the domain of semi-solid glass AM. This is possible because during AM (of semi-solid glass), any extruded glass experiences brief transit times through the nozzle—deliberately minimising exposure to temperature ranges associated with devitrification. This can possibly enable the processing of waste glass with fewer conventional challenges. However, again, further research is warranted to better understand glass recyclability through the lens of increasingly circular economies.

3.3.2. The Production of Glass Filament

To fabricate glass filament for high-viscosity methods, glass is typically ‘pulled’ mechanically from a bulk mass into a rod. This process is often given different (and essentially interchangeable) names in the glass art community, with the glass rod being known as a cane, stringers, or filament. Filament production may be performed in numerous ways. Traditionally, the Venetian style of glass cane production involves two people ‘hand-pulling’ a mass of glass in opposite directions, to thin the glass out into strands. Hand-pulled filament is compatible with some glass 3D printers; however, for precise additive manufacturing, it remains crucial to utilise filament that is consistent in size, to allow the accurate control of flow rates and extruded geometry. High-tolerance glass filament can be difficult to fabricate with manual techniques. Instead, draw towers and vitrigraph kiln [94] systems may be used to produce high-quality glass filament suited for additive manufacturing. A relatively recent instrument known as the MGP Vitri-Glass^TM, was designed with the primary application of recycling glass and producing precision filament for glass AM, as depicted in Figure 24. In addition to dimensional accuracy, the material quality of filaments is a consideration. As filaments are directly extruded, impurities within the filament such as air bubbles or foreign materials will be present in the final printed object.

3.4. Postprocessing of AM Glass

To date, comparatively little exploration of the postprocessing of additively manufactured glass has been reported in the open literature. A small number of papers have documented cutting and polishing objects produced by glass AM to further characterise the object [10,22,31,67,68]. While the polishing process was not the area of investigation for these studies, it was still shown to yield some improvement in optical quality, reducing the visibility of the layer lines which distort the surface in a typical AM process. Kiln firing has also been attempted as a starting point for the postprocessing of layer lines. Many AM glass techniques involve the use of a heated chamber, annealing the produced object at the onset of printing and extending it after the process is completed [10,88,89]. Increasing the chamber temperature beyond the annealing range initiates a fusing and slumping process which reduces the visibility of the layer lines, at the cost of precise object geometry. An example of a 3D-printed and re-fired glass component via the use of a kiln is depicted in Figure 25. Investigation directly aimed at these methods of postprocessing is needed to characterise the complexities between layer line visibility, the resolution of object geometry, and ease of application.

As an extension to kiln fusing and slumping, fibre optic applications also offer a specialised avenue for the postprocessing of glass AM objects. There have been multiple studies which demonstrate the successful drawing of fibre from a glass preform produced by AM [50,65,66,67,68,82,98,99,100]. In these studies, an AM process constructs glass into a complex preform geometry, which is subsequently heated to a suitable fusing temperature where it may be drawn into a fibre, free from layer lines and other artefacts of the AM process.

In an examination of postprocessing through a more structural lens, hot working has the potential to further expand the extent of applications. An analogous concept is the welding of polymeric or metallic components. Such a procedure expands the possible size of final components or provides an opportunity to create a unique object geometry. The (seamless) joining of glass printed parts would have a particular impact on large-scale applications such as architecture/building. In the context of glass postprocessing, hot working can consist of re-firing 3D-printed glass with the use of a kiln, or through open-air glassblowing/forming techniques.

The postprocessing of 3D-printed glass components utilising glassblowing techniques entails the manipulation of glass material at temperatures (and the associated viscosities) conducive to both sculpting and fusion. To date, demonstrations of this process have been seen in artistic exhibits or showcases in person. One such demonstration took place at AusGlass 2023, where six glass 3D-printed boats were amalgamated into a singular glassblown vessel, by glass artists James Devereux and Louis Thompson. Similarly, at the Glass Art Society conference (2023), glass artist Yazmin Dababneh revealed the integration of glass 3D prints to form a goblet, depicted in Figure 26. Across such demonstrations of postprocessing, it has been observed that 3D-printed glass parts exhibit performance characteristics akin to traditional glass materials (with amenability to conventional forming). This is promising for the integration of methods, whilst indicating that further research is warranted to expand upon the knowledge base within this domain. In another example of postprocessing, the cutting and fusion of 3D-printed glass with a kiln is demonstrated in Figure 27.

The removal of any support material (which is relevant in the case of, say, EAM methods) may also be considered postprocessing. Support material is typically employed during 3D printing to serve as temporary scaffolding, to support overhangs or bridges, where a complex design has geometry inclusive of angular components greater than ~45 degrees. To easily remove support material once the print is complete, an offset can be introduced between the printed component and the support. This helps to ensure an interlayer bond strength that is weaker than that of the other (permanent) layers. An example of this is provided in Figure 28 for glass AM using EAM, where an offset of 1.5 mm was adequate to allow the simple removal of support material (commonly carried out manually/mechanically with pliers). The removal of support material often results in a surface finish that is suboptimal, both optically and mechanically. Further printing parameter optimisation (iterating upon offset, width of deposition, material temperature and cooling, etc.) may improve the surface finish of the support material interface, or one of the previous postprocessing methods may be applied.

4. Properties of AM Glass

4.1. Mechanical Properties

To advance the field of glass AM technologies, a comprehensive characterisation and understanding of the mechanical properties of AM-prepared glass (using the many AM methods described herein) will be critical. Mechanical testing is fundamental, offering valuable information on load bearing capacity, fracture-related performance, and the overall reliability of manufactured components. This is essential for both ensuring product quality (of glass AM components) and meeting industry standards (in relation to the presently known performance of glasses in industrial use). Knowledge of mechanical properties enables informed material selection and design optimisation, which leads to waste reduction and enhanced overall performance. This can be achieved through assessing how AM glass behaves under various loading conditions, contributing to the evolution of glass AM technology, and realising the full potential of structural (or any new) applications.

As with several facets of the domain of glass AM, there is a paucity of mechanical test data on glass AM parts. Early investigations, such as the original testing conducted by Klein in 2015 [10], focused on semi-solid processed glass (using an approach that is also popularised as the MIT method). In the study, two sample types were compared, cut from a larger manufactured sample. One set was printed in a heated chamber set to 482 °C, and the other was printed with the chamber at room temperature (25 °C). Following an annealing cycle, both sample types exhibited negligible stress concentration along the layers. However, radial stresses were observed within the layers of the room-temperature samples. A three-point bending loading scenario was employed, applying a load along the glass layers (presented in Table 4). This loading orientation was selected to investigate interlayer adhesion. Of note is that samples printed in a chamber at 482 °C demonstrated a flexural strength five times greater than the room-temperature samples. This was attributed to the temperature differences during layer deposition. A large temperature difference between the deposited layer and previous layer appears more likely to impact (and potentially weaken) interlayer bonding. Additionally, the study noted that the fracture routes of 3D-printed samples prepared at 482 °C were observed to have cracks without a precise trajectory.

More rigorous versions of this study were performed by Inamura in 2018 [15,64], where the peak stress in samples was quantified for tensile and compressive loading scenarios. Notable was the significant decrease in compressive strength from the small-scale rectangular samples to the complete large-scale prints. The cause of this was believed to be influences on the boundary conditions (i.e., in-plane friction) causing out-of-plane bending moments and compound stresses at the perimeters. This highlights that the characterisation of such influences is of great importance to the design of large-scale glass objects. Art and architecture necessitate complex forms which will experience loading scenarios much more complicated than those seen in a laboratory.

In another study, Seel (2018) [24] focused on AM glass structures that were deposited directly onto glass plates. In the study, ‘joint strength’ (akin to adhesion strength) was investigated, which is a fundamental building block for glass AM. Borosilicate glass rods were fused to borosilicate glass plates prior to annealing. Testing was then carried out in a cantilever bending orientation (Table 4). The results indicated that ~66% of fractures occurred at the site of the joint. In comparing mean values from experimental testing, it was noted that failures in the rod (as opposed to the joint region) tended to show higher failure stresses. The study indicated that the joint region is potentially a weak point, a concept that is not isolated to glass AM, but an important consideration for the AM of any material. It is, however, noted that similar studies are yet to be carried out in the open literature.

More recently, Chaddeh (2023) [102] further explored glass structures upon sheet glass; however, in the study, a shearing test was used, and the material explored was fused silica glass (Table 4). It was noted that in the majority of test cases, the failure occurred in the plate, leaving the joint intact. As the study is one of the only ones to explore a glass material that is closer to a commodity version of glass, the results indicate promise for the reliable use of AM glass components.

Most recently, a comprehensive study was undertaken by Nowak (2024) [103], who explored the mechanical behaviour of 3D-printed glass ‘dog-bone’ samples, tested to failure in a 4-pt bending loading configuration (Table 4). In all, fifteen samples, categorised into three distinct infill orientations, were produced, and glass AM prepared by EAM as a function of build orientation, namely, 0°, 45°, and 90° orientations, was explored. The results (reported in Table 4) notably reveal that the 90° specimens exhibited a mean stress value approximately 44% higher than that of its counterpart specimens (i.e., the 0°and 45° specimens). This disparity was posited to be associated with the toolpath required to produce each infill orientation. The 90° configuration is characterised by a more ‘streamlined’ toolpath, while the 0° orientation necessitates long travel movements. Additionally, the prolonged depositions inherent in the 0° samples require the tool head to deviate from freshly deposited layers, towards the comparatively cooler layers. As a consequence of these findings from EAM specimens prepared with the MGP process, there appears to be a need for important future (fundamental) work regarding the quantification of different thermal profiles generated from glass AM toolpaths—which are likely to influence interlayer bond strength. It is of merit to note that the EAM specimens (in any configuration) perform rather impressively relative to conventionally prepared soda–lime glass, suggesting that glass AM does not degrade glass properties.

These results suggest that interlayer bond strength in glass 3D printing depends on nozzle temperature, environment temperature, and print settings such as speed and layer size. Higher temperatures during glass deposition have been shown to improve part strength. While delamination in glass AM has not been reported, further research is needed to quantify its occurrence under different process parameters.

Table 4. Summary of mechanical testing conducted on AM glass samples.

AM Process	Glass Material	Loading Scenario	Sample Dimensions	Failure Stress (N/mm2)	Image	Ref.
Low-viscosity glass AM	Soda–Lime	3-pt Bend	10.8 × 17.7 × 56.1 mm	Not Provided		[10]
Low-viscosity glass AM	Soda–Lime	3-pt Bend	~70 × 13.1 × 8.5 mm (annealed)	41		[64]
Low-viscosity glass AM	Soda–Lime	3-pt Bend	~80 × 13.7 × 9 mm (annealed + chemically tempered)	41		[64]
Low-viscosity glass AM	Soda–Lime	3-pt Bend	100 × 13 × 9 mm	41 ± 15		[15]
Low-viscosity glass AM	Soda–Lime	Compressive Loading	40 × 13 × 9 mm	254 ± 23		[15]
Manual Fusing	Borosilicate	Cantilever Beam Bending Test	Rod: 3 mm diam. Plate: 3.4 mm thick	114.6		[24]
DED	Silica	Shear Test	Plate: 50.2 × 24.9 × 3.2 mm AM deposit: ~6 mm diameter bead, 23 mm length	34.5		[102]
SLM	Soda–Lime	3-pt Bend	1 × 2 × 8 mm	~6.5		[42]
Light-Based	Silica	3-pt Bend	Truss structure: ~2 × 2 × 4 mm	187.7		[84]
EAM	Soda–Lime	4-pt Bend	150 × 40 × 3 mm 0° infill angle	40		[103]
EAM	Soda–Lime	4-pt Bend	150 × 40 × 3 mm 45° infill angle	39		[103]
EAM	Soda–Lime	4-pt Bend	150 × 40 × 3 mm 90° infill angle	56		[103]

Table 4. Summary of mechanical testing conducted on AM glass samples.

AM Process	Glass Material	Loading Scenario	Sample Dimensions	Failure Stress (N/mm2)	Image	Ref.
Low-viscosity glass AM	Soda–Lime	3-pt Bend	10.8 × 17.7 × 56.1 mm	Not Provided		[10]
Low-viscosity glass AM	Soda–Lime	3-pt Bend	~70 × 13.1 × 8.5 mm (annealed)	41		[64]
Low-viscosity glass AM	Soda–Lime	3-pt Bend	~80 × 13.7 × 9 mm (annealed + chemically tempered)	41		[64]
Low-viscosity glass AM	Soda–Lime	3-pt Bend	100 × 13 × 9 mm	41 ± 15		[15]
Low-viscosity glass AM	Soda–Lime	Compressive Loading	40 × 13 × 9 mm	254 ± 23		[15]
Manual Fusing	Borosilicate	Cantilever Beam Bending Test	Rod: 3 mm diam. Plate: 3.4 mm thick	114.6		[24]
DED	Silica	Shear Test	Plate: 50.2 × 24.9 × 3.2 mm AM deposit: ~6 mm diameter bead, 23 mm length	34.5		[102]
SLM	Soda–Lime	3-pt Bend	1 × 2 × 8 mm	~6.5		[42]
Light-Based	Silica	3-pt Bend	Truss structure: ~2 × 2 × 4 mm	187.7		[84]
EAM	Soda–Lime	4-pt Bend	150 × 40 × 3 mm 0° infill angle	40		[103]
EAM	Soda–Lime	4-pt Bend	150 × 40 × 3 mm 45° infill angle	39		[103]
EAM	Soda–Lime	4-pt Bend	150 × 40 × 3 mm 90° infill angle	56		[103]

4.2. Thermal Properties

For meaningful improvements in direct processing methods for glass AM, a clear need exists for investigations regarding a detailed understanding of thermal gradients. Discussed in Section 1.2, thermal gradients formed while printing may create stresses leading to fracture. The characterisation of such gradients can be applied to advise improvements in object scale and complexity, by augmenting the existing methods with more intelligent process control.

For example, while a heated chamber has enabled the production of glass objects without fracture, there will ultimately be limitations in scale with this method. Exceeding a certain size of object, an enclosing heated chamber will become problematic due to cost, heating uniformity, and user safety. An alternative approach may be derived from detailed knowledge of cooling and the resulting temperature within a printed model. Thermal fracture may be avoided by intelligently accounting for thermal gradients and introducing self-annealing behaviour, possibly by controlling the rate and size of material deposition.

Similarly, some methods which enable complex geometries in polymer AM (bridging and overhangs) may be introduced based on deeper knowledge of cooling behaviour. Glass AM already has the advantage of being able to exploit wide ranges of glass viscosity—even allowing some non-planar printing, as observed in Figure 29. Bridging and overhangs, however, are an issue for some methods as they require material deposition to occur in the absence of supporting material from previous layers. Precise knowledge of viscosity and the cooling rate is required to develop a bridging function which deposits material without sagging and with the capability of supporting subsequent layers.

The study of Klein et al. [10] contains some preliminary investigation into problem behaviours during printing, using thermal imaging to quantify a temperature gradient between layers. Iteration upon this known gradient is, however, ignored in favour of a heated chamber to prevent thermal fracture. Essential further research would build upon this in situ monitoring, manufacturing objects of a known, standardised geometry to compare the thermal behaviours of various materials and process parameters. For example, the cross-section size of deposition (influenced by temperature, feed rate, and layer height) may be linked to the rate of cooling per deposited material. Applied to more complex object geometry, knowledge of cross-section cooling may advise areas of printing with problematic thermal gradients and allow more suitable parameters to be chosen. For example, the size of material deposition may be increased in a certain area, retaining temperature for a longer period and reducing the severity of the thermal gradient.

An alternative, but equally viable, approach to understand and optimise thermal behaviours is simulation. The modelling of the true thermal behaviours in glass is a complex subject. Glass materials are partially transparent, making them what is known as a participating medium. In cooling, thermal modelling must therefore account for the behaviour of conduction, coupled with radiation transfer throughout the material [104,105,106]. Such modelling has many implementations in modern computational fluid dynamics. ANSYS Fluent, for example, may be configured to apply a finite volume approach, approximating radiation with a discrete ordinates method (ANSYS documentation [107] cites [108,109,110] as the basis for modelling). There is vital work to be conducted in applying this software to scenarios specific to glass AM. The examination of isolated mechanisms, such as the deposition of a single layer upon another, will provide insight into the complex heat transfer and viscoelastic deformation in the process. Lou et al. have begun this investigation for a DED process [12], applying ANSYS Fluent to the melting of a single track of borosilicate glass. Similar research would benefit the SLM and semi-solid processing methods, informing the interaction of hot, deposited material with cooled previous layers.

The holistic modelling of an entire object during production is also essential to build upon the simulation and monitoring of the isolated mechanisms. In direct methods, material deposition traces a heat source across the object as it is produced. Certain printed geometries may exacerbate the temperature gradient created by this heat source. There will be a critical interplay between certain object geometries (long meandering paths, thin and tall sections, etc.) and process parameters (feed rate, cooling of deposition, etc.) that must be understood for the stable production of objects without thermal fracture.

It is likely that the discussed knowledge must also be applied outside of academia, where computational fluid dynamics is presently too complex for practical use in printing software (e.g., slicing software). This software is required to perform rapid calculations and be configured without specialised knowledge. A simplified model is needed for this type of implementation. The exploration of both empirical (e.g., in situ temperature monitoring) and simulated data (e.g., from computational fluid dynamics) of transient temperature in a specific glass AM context may provide some insight into the construction of this model, which would also be significantly enabled by machine learning.

4.3. Optical Properties

The optical properties of AM glass have been a primary interest for the direct ink writing, light-based, SLM, and DED methods [3]. This is due to the high resolution achieved with these techniques, with a focus on optical applications such as lenses and fibres. An AM process unlocks a new level of complexity for lenses and fibre preforms, allowing the precise control of geometry and even the possibility of varying material (and the associated refractive index) across a printed object [50,56,60,66,67,68,79,82,83,84,98,111]. However, for glass EAM processes, optical properties have been less studied. This is because the resolution achieved using these techniques is more suited for alternative applications, such as architecture, art, prototyping, packaging, and so forth. Common interest does revolve around the layer lines and whether these can be altered to improve the optical quality of an object produced with glass EAM. Polishing and firing methods have been mentioned in Section 3.4 which show that optical properties can be altered. When it comes to exploring new forms and functions, there is also the possibility of exploiting the layer lines to achieve an aesthetic. For example, their unique interaction with light may be used to create caustic patterns in combination with a light source, such as those depicted earlier in Figure 14. More research is required to make definitive statements on the optical properties of additively manufactured glass.

5. Summary and Future Needs

5.1. Summary

The review herein has surveyed a body of work, most of which consists of relatively recent studies, which indicate that the 3D printing of glass continues to progress significantly—and appears to be here to stay. The processes explored reveal associated relevance for demonstrating considerable potential in various industries, such as architecture, art, optics, and healthcare. The additive manufacturing of glass has, to date, employed approaches and methods which are relevant across a very broad range of length scales (from microns to many tens of centimetres).

Innovations in this field are broad and are associated with unique methodologies which are being employed for the AM of glass. For example, the EAM process with semi-solid glass processing (both at high and low viscosity) has transitioned from research settings to commercial uptake. Binder-based methods, which involve glass feedstock held in a green object that is later fused, has also seen broad uptake. Uniquely, even laser-based methods are being explored, with a varying degree of success as they remain in the R&D phase. Researchers are also exploring hybrid techniques combining 3D printing with traditional glassmaking processes to enhance the properties and aesthetic qualities of the final products. Of the works reviewed herein, the variety of glasses that have been utilised in glass AM to date has been confined to a very small subset of glass chemistries (of the thousands of available glasses explored). However, recycled soda–lime glass has been one of the most widely explored materials—which augurs well from a sustainability perspective. Despite some of the challenges expounded in this review (and future work/needs elaborated below), the potential for customised, complex structures makes glass AM a critical future technology. Geometry control and the emergence of metamaterials through generative design will allow the preparation of glass structures by AM for a potentially wide variety of new applications for glass.

5.2. Challenges and Future Needs

Some of the key questions or future directions regarding glass AM that have become evident from the review herein include the following:

• There is a need for research to develop the scalability and reproducibility of glass AM methods (and associated feedstock material processing) to produce models/components at a larger length scale (i.e., the exploration of glass AM which incorporates one (or more) dimensions that are >300 mm);
• Depending on the method of glass AM employed, achieving (and maintaining) the optical clarity of printed glass objects is a concern for applications that require clarity or complete transparency. This limitation is one that is not, however, specific to glass AM, but an existing limitation of all AM methods (including polymeric EAM, and metal and ceramic AM), with the possible exception of resin-based stereolithography methods;
• Further research in and analysis of postprocessing methods used upon glass AM-produced objects are required. The available information is insufficient in this area, with only incidental exploration of polishing and annealing or artistic techniques documented in the literature. Further exploration of more technical postprocessing presents an opportunity to improve aesthetic, optical, and mechanical properties in parallel with direct process improvements in glass AM;
• Studies should demonstrate the ability to exploit design limits specific to glass printing, such as the overhang angle, resolution of intricacy, capacity for complex topology, and integration of complex features. These have not been demonstrated to date, and have not been correlated with computational models;
• There is a need for a better understanding (and modelling) of the strength of glass in the context of glass AM. Presently, there is a paucity of data in the domain of glass AM with respect to mechanical properties. There is also additional subtlety in mechanical properties across length scales. The performance of full, large-scale elements and objects is essential for art and architecture but is yet to be tested. Whilst component strength is critical, there is also a need to better understand the possibility of interlayer delamination in glass AM, which will require targeted testing;
• Further investigation into the detailed thermal behaviours of deposited glass during printing needs to be performed. Detailed knowledge of thermal behaviour does not exist for the specific area of glass 3D printing, limiting progress with the poor scalability of current implemented methods. The simulation of thermal behaviours and in situ thermal monitoring will provide vital insight, forming a basis for advanced process parameter optimisation;
• Commercial products prepared by glass AM remain in their infancy, and the ongoing (and broader) use of glass AM technology will illuminate further opportunities and prospects. Aspects similar to those in the existing glass industry must be investigated, such as the quality control procedures used for mass-produced glass items, including the reproducibility and consistency of properties;
• There are prospects for machine learning to contribute towards the rapid optimisation of glass AM processing parameters, allowing the accelerated insertion of glass AM technology in commercial products.