1. Introduction

3D printing – or additive manufacturing (AM) - of solid objects from a digital file has inspired what is being referred to as a digital revolution in manufacturing or “Industry 4.0” [1, 2]. Starting in the late 1970’s, research in the area of freeform fabrication of material led to a rapid prototyping capability and eventually to commercialization of mostly polymer-based systems in the 1980’s that enabled designers to iterate quickly through new designs. However, the use of 3D printing technologies for widely deployable short run manufacturing would not be realized until decades later with the advent of low cost subsystems for metal AM. In particular, low cost and efficient fiber lasers [3] led to wider spread use of powder bed fusion systems capable of printing industrially-relevant alloys such as Ti-6Al-4V, steel 316L, AlSi10Mg and nickel superalloys [4–7]. Guided by ever more sophisticated modeling efforts [8, 9], metal AM machines are now seeing real world applications in the aerospace, automotive, healthcare and jewelry industries. In the latter two industries, the relevant part length scales and volumes suggest modest manufacturing throughput since personalization drives demand. However, in heavy industries such as automotive and aerospace, mass production, larger length scales or both require that metal AM technologies must scale up accordingly. Unfortunately, because a raster-scanned source (either with a laser or an electron beam) is at the heart of today’s metal AM approaches, the print times scales linearly with volume as roughly 100 cc/hour (for steel 316L) for a single 1 kW beam scanning at ~1 m/s. Multiple beams can be added to parallelize the process, or the beam intensity/scan speed can be increased, but systems quickly become complex even after 4 beams, and eventually scan speed limitations due to microstructural defect creation are difficult to overcome.

A more elegant and scalable extension of multiple laser beams could be achieved by spatially shaping a single, wide-area high power beam as was recently demonstrated at the National Ignition Facility (NIF) at Lawrence Livermore National Laboratory [ref Heebner]. Beam shaping in NIF is achieved using an Optically-Addressable Light Valve (OALV), and is performed for the purpose of reducing the intensity of portions of the beam which would otherwise cause or exacerbate laser damage on certain optics. The ability to temporarily shadow, or “block” these few isolated sites from high fluence laser pulses has enhanced operational flexibility by enabling uninterrupted use of NIF at near peak laser performance until such time as the optic can be removed, repaired offline either through recycling or refinishing, and replaced [10]. However, the application and requirements of an OALV for laser materials processing, e.g. additive manufacturing, has not been explored.

Here we propose an alternative method for creating additively manufactured parts using a wide-area, optically-addressable photomask supplied with a hybrid light source consisting of a low cost diode laser array and a Q-switched pulsed laser. Since the majority of the heat input is supplied by the diode laser component, our system can be shown to scale more effectively as opposed to the more costly fiber laser-based counterparts even with the addition of the pulsed laser source. The novel use of an OALV [11, 12], not only allows addressing on and off regions of the build but also offers the potential to ‘grayscale’ an image providing selective thermal gradients to be imposed to control residual stress and microstructure.

2. Description of laser system

Figure 1 displays the optical systems used for layer-by-layer printing using our hybrid two laser, OALV-based approach. Figure 1(a) shows the Diode-based Additive Manufacturing (DiAM) laser system which is comprised of a set of optical subsystems. These subsystems are a set of diode lasers, an incoherent beam combining optical system, a short pulse laser and delivery optics, the OALV, image projector, and image delivery optics to the print plane. Figure 1(b) shows the layered structure of the OALV and illustrates its basic functionality. From left to right, the OALV consists of a BK7 glass substrate, a layer of conductive indium tin oxide (ITO), a short gap defined by spacer spheres, a twisted nematic liquid crystal layer, an additional spacer gap, a photoconductive Bismuth Silicon Oxide (BSO) layer, and a final conductive ITO layer. When 470 nm light from an incoherent projector is incident on the photoconductive layer, there is a local short to the BSO layer causing the LC to align with the applied field, thus disabling polarization rotation. Additional details regarding the OALV device performance are discussed below.

 

Fig. 1 (a) Diode-based Additive Manufacturing (DiAM) optical layout. Four 1007 nm, high power, laser diode arrays (~1.25 kW each) are spatially-multiplexed through a set of turning mirrors and a cylindrical lens (L1) before being re-imaged through a pair of lenses (L2, L3) and directed into a beam homogenizer. Flat-fielded light then continues through a second set of telescopic re-imaging optics (L4, L5) and on to a polarizer-compensator pair before being temporally- and wavelength-multiplexed with a 7 ns pulsed 1064 nm laser beam using a (IR) dichroic mirror. This horizontally-polarized hybrid IR laser pulse then passes through the optically-addressable light valve (OALV) where a 470 nm patterned light emitting diode (LED) beam selectively addresses a polarization (non-)rotation in the IR beam: IR light that is addressed as ‘on’ pixel passes through both the OALV and subsequent polarizer with a horizontal polarization while IR light that is unaddressed as ‘off’ pixels experience a polarization rotation to vertical at the OALV and is directed out of the optical system with a polarizer and into a beam dump. The remaining patterned IR beam is then directed onto the build plane via turning mirrors and imaging optics (L6-L8) to print a metal layer at once by melting the powder bed. (b) Expanded view of the OALV displaying the individual components responsible for selectively addressing the high power hybrid laser beam. “AR_x” refers to anti-reflection coatings. Note that for clarity, the 470 nm LED light source in (b) is displayed as counter-propagating relative to the 1007/1064 nm beam; in the actual system, both 1007 and 470 nm beams are co-propagating as shown in (a).

Download Full Size | PDF

The laser diode sources are made up of a set of four diode arrays (60 individual bars each; Trumpf, Germany) and can produce 1.25 kW each in continuous wave (CW) operation, 4.8 kW total at 150 A. The laser output can be pulsed with a width as short as 80 μs at full peak intensity with a center wavelength of 1007 nm and a +/−3 nm FWHM bandwidth. A set of turning mirrors and lenses spatially multiplexes the output of the four diode arrays onto a homogenizer. The homogenizer is a hollow rectangular cuboid with gold plating along the inner surface. The purpose of the homogenizer is to take a non-uniform tiled beam input and convert it to a uniform square output with a large number of modes (low speckle size) in a rectangular 6x6 mm² shape. Following the homogenizer is a polarizer and a set of relay optics that image the homogenizer output onto the OALV with a de-magnification ratio of 2.9. This results in a uniform intensity 18x18 mm² square spot at the OALV. In addition to the diodes, a 7 ns short pulse beam from a Q-switched Nd:YAG laser (Continuum, USA) is used to finally initiate the process and melt the powder. A custom powder spreader was used for DiAM builds. 1” pistons are adjusted using motorize lab jacks that have better than 1 μm repeatability. The recoater is mounted on a translation stage with a maximum velocity of 200 mm/s. The build piston is moved down 50 μm to set the layer thickness before recoating and at the end of printing is lifted up to remove the substrate with the printed part attached to it. The entire spreading mechanism is placed in a standard glove box that is purged with ultra-high purity (99.999%) argon. Sn powders were acquired from Goodfellow, USA.

In what is presented here, the OALV was used in a binary-on/off configuration. However, as shown below and described previously [13], OALVs can be easily operated in grayscale-mode by appropriately scaling the intensity in each pixel of the 470 nm projector image. The OALV transmission versus incident 470 nm intensity depends on the voltage and frequency at which the OALV is driven. When operating in the binary configuration it is desirable to use settings that produce a sharp onset of IR transmission as opposed to a gradual rise with input intensity. One major advantage of using an OALV as a beam shaper for additive manufacturing is that the input beam does not need to be single mode or low divergence. Unlike alternate methods which rely on scanning a beam with a small spot size along the print plane, this method can generate an entire print layer in one pulse using lasers with high divergence and a large number of modes. Laser diode arrays with linearly polarized output are well-suited for this task.

The primary advantage of using low cost diodes to selectively melt material is the ease in scaling up to larger areas and correspondingly higher powers using spatial or wavelength multiplexing of sources. However, when considering the use of this technique in material processing of micron-scale layers of powders, the temporal characteristic of the incident beam and thermal properties of the material must be considered. For a 20 ms pulse and a thermal diffusivity D~0.11 cm²/s, the thermal diffusion length in Sn powder is approximately given by L $=2\sqrt{D\tau }$ ~0.94 mm which would negate the fine feature scale afforded by the optical resolution of the system in the x-y plane and the powder particle size in the z-axis. In order to precisely modify the energy deposition to limit the extent of material phase change, we introduce a second laser source with a pulse length τ_s << L²/4D. This additional short-pulse beam is generated from the 7 ns Nd:YAG laser source with a maximum energy of 2 J at 1064 nm and is combined with the 1007 nm diode beam using a dichroic filter just prior to the OALV (see Fig. 1). Finally, a projector image at 470 nm is combined with the two laser beams using another dichroic mirror. The projector image and the image of the homogenizer output are coincident on the OALV. The OALV rotates the polarization of the “off” pixels, while the “on” pixels remain in their original orientation. A final polarizer after the OALV rejects the light corresponding to the “off” pixels while allowing the light corresponding to the “on” pixels to pass through to the final image relay optics and the build plane.

3. OALV operation and performance

Conventional liquid crystal (LC) light modulators such as those found in LC-based displays or projectors generally consist of a 2D array of discrete LC cells or pixels. The transmission of each pixel is addressed on an individual basis through a matrix of applied voltages. By contrast, OALVs based on liquid crystals consist of a single, large liquid crystal cell that is addressed by a 2D light image. The local intensity of the address image controls the local transmission of a patch of the OALV. This addressing mechanism is achieved with a photoconductive layer inserted in series with the LC layer, both surrounded by electrodes that provide a common voltage across the entire device. Illuminating the photoconductive layer with a light image above the bandgap of the photoconductor locally shorts patches where high intensity is present enabling the voltage to be applied locally across the liquid crystal layer. Where no address light is present, the voltage resides across the high impedance photoconductive layer. Because of this unique property, LC OALVs have been used for niche applications including light-by-light switching, ultrafast pulse shaping, and laser beam shaping [14–17].

Optical addressing circumvents the need to fabricate a pixelated matrix with electrical backplane. Through this approach, the auxiliary addressed image can be patterned with a conventional, pixelated electrically addressed modulator. This neatly divides the challenges into two subcomponents: 1) auxiliary images can be created at low power, low coherence, and can leverage any of a number of commercial LC arrays and 2) the OALV does not require an electrical backplane and hence can be scaled to large apertures and handle high powers [15]. An example light pattern based on an image resolution target is shown in Fig. 2. The OALV used in this system implements BSO as the photoconductor which is addressed at 470 nm and modulates the amplitude of the diode light (through polarization rotation). It can support a 50 μm spatial resolution at the OALV image plane, 10 ms refresh rate, and up to 100:1 extinction ratio over a 22 mm × 34 mm clear aperture. Figure 2 shows the image resolution target at three planes of interest for comparison. The digital image is sent to the projector, as shown in Fig. 2(a). The projector image is relayed to the light valve plane (b) where the image projector light (470 nm), diode beam (1007 nm), and short pulse beam (1064 nm) all intersect. As described above, the liquid crystal region of the OALV rotates the polarization of the incoming IR beams when there is no incident 470 nm light on a particular pixel. The rotated beam is then rejected by a subsequent polarizer with an efficiency of 95%. However, the OALV responds to the 470 nm “on” regions by preventing rotation, allowing the IR beams to locally pass through the polarizer and propagate the image to the build plane shown in Fig. 2(c). The imaging target shown in Fig. 2(a) is scaled at the projector and undergoes a slight magnification of 0.85 at the relay to the OALV plane, finally being further magnified by 0.32 as compared to the original image size in the relay to the print plane. This target shows there are clearly discernable features in the OALV plane down to less than 40 µm (e.g. group 3, element 5), and features in the print plane down to less than 50 µm (group 1, element 6). For comparison, typical SLM systems operate with raster-scanning beam diameters in the range of 50-150 μm depending on the build chamber height and aperture size of the last optical element (which roughly determines the numerical aperture of the system).

 

Fig. 2 Comparison of (a) original digital (binary) file, (b) 470 nm wavelength projected light pattern incident on the OALV and (c) ~1 μm wavelength hybrid beam pattern projected onto the sample plane (false color). Images shown in (b) and (c) were collected at the OALV and sample planes respectively using a 12-bit digital camera. (d) Transmission and extinction (1/leakage) performance of the OALV, demonstrating the optical response and potential for grayscale operation at intermediate 470 nm intensities, e.g. 0-10 and ~5-35 mW/cm² for 500 and 1500 Hz drive frequencies respectively.

Download Full Size | PDF

Quantitative contrast performance measurements are shown in Fig. 2(d) for two system configurations. Both measurements were taken with the OALV driven at 30 V, but the frequency was changed from 500 Hz, producing a sharp transmission rise from 0 to ~96%, to 1500 Hz, producing a more gradual increase from 0 to ~88%. Thus, the OALV can be frequency-tuned to adjust device response such that a more sensitive variation of transmission versus the projector light can be achieved. This grayscale adjustment could be useful for controlling residual stress or microstructure of a build based on tailored laser intensity profiles which can drive thermal gradients [18]. The data was collected by fixing an aperture immediately before the OALV so that the area is known. A variable attenuator was placed directly after the projector, and the total power through the aperture was calibrated to a power meter placed behind the “VisIR” dichroic mirror to measure the leaked through light. As shown in Fig. 2(d), this allowed the measurement of IR transmission versus intensity at 470 nm and the determination of the OALV extinction ratio which is shown to vary from 0 to ~21 at 500 Hz.

Although the OALVs used in this study were originally designed for and implemented in the National Ignition Facility’s laser front end [15], the advantages of using an OALV for layer-by-layer printing with high power diode patterning are many fold. First and foremost this introduces the ability to scale to large sizes limited only by the size of the photoconductive layer in the OALV and beam relay optics. Furthermore, because all components (LC, ITO, BSO, borosilicate glass) have low optical absorption coefficients, the integrated system can have very high optical damage resistance. Imaging quality benefits from non-pixelated light shaping which minimizes artifacts associated with discrete pixels present in other spatial light modulators. Finally, the gray-scaling possible with OALVs allows tailored light patterns to be used which can tune thermal gradients and cooling rates.

4. Roughness as a function of laser fluences

To demonstrate the dual beam operation and efficacy of melting metal powders, we studied the dependence of the final surface roughness on the incident laser energies from the two lasers. Height variations greater than ~2 × the layer thickness of each layer during a print is known to affect final part quality through pore generation [19] and surface roughness [20]. The physics driving morphology of melted powder layers is still a subject of active research [21, 22], but is known to involve two primary effects: keyhole-mode evaporation and incomplete melting of powder particles. In the former case, temperatures at the melt surface exceed the boiling point of the material and cause vapor pressures that drive splashing and pore formation. For the latter, particles may either resist complete melting under direct laser illumination or fall onto semi-molten tracks following laser irradiation after being transported to the processing area from gas entrainment. These effects can be exacerbated upon subsequent layer processing due to thick or thin layers caused by low and high regions in the powder. Low regions in the next to last layer result in thicker powder during welding of the last layer. This can lead to further incomplete melting between the last two layers. Conversely, high regions in the next to last layer result in thin areas of powder during welding of the last layer. This can lead to overheating and porosity from boiling. Each layer is dependent on the layer before it and roughness can be built up over many layers in the printing process. Finally, if the processing parameters are unable to completely melt large areas of powder, the layer may not bond to the previous layer, causing the layer to be swept away during recoating resulting in loss of the print, or a crash of the coater blade leading to build failure.

To assess these issues we measured the surface roughness as a function of the system parameters. As a first demonstration, we chose Sn powder due to the low melting point and therefore minimal laser intensity on the OALV. Spherically-shaped Sn powder was cast into an ingot using a crucible and then machined into a substrate plate 1” in diameter and 10 mm thick. Subsequent samples of the same tin source powder were spread across the substrate plate using our experimental spreading system. The single layer powder uniformity can be seen in the top panel of Fig. 3 with an average thickness of 60 μm. As the layer is made thinner, the larger particles create long drag marks, which can be a source of defect formation and eventual surface roughness. Some variation in layer thickness was present, as the powder layer tended to be thinner and more sparse at the side of the substrate where the spreader first crosses and deeper as it continues across the substrate. The powder layer is illuminated by the hybrid laser source yielding a fluence range that produced permanent deposition. The lower side of the range at ~20 J/cm² for the long pulse laser and 1.5 J/cm² for the short pulse laser was chosen based on the minimum for particle melting. The high fluence side of the long pulse range at ~35 J/cm² was chosen by excessive boiling which would cause splatter of the liquid melted onto nearby sites. The high side of the short pulse range at ~2 J/cm² was chosen to fall within the laser damage initiation bounds of the optical delivery system, including that of the OALV.

 

Fig. 3 Confocal scanning laser profilometry of processed powder layers. As shown, a row of 500 μm diameter laser weld spots are visible, before and after the metal powder is removed. The short pulse energy ranged from ~1.8 to 2.6 J/cm² from left to right in 0.5 J/cm² increments while the longer diode pulse was fixed at ~45 J/cm². Weld spots initially visible from 0 to 7 mm along the x-axis did not adhere to the surface upon removal of the residual powder and are omitted in the after image.

Download Full Size | PDF

After printing a single layer of three parameter sweeps extending across the substrate, it was discovered that bonding of the melted powder layer to the substrate was influenced by the morphology of the powder. In the top panel of Fig. 3, 15 individual deposition sites in three groups of 5 were created, where it can be observed (in the bottom panel) that after removing the powder, only 7 of the sites were welded to the substrate. These sites are all in regions where drag features through the powder resulted in some of the substrate being directly exposed to the laser. One explanation for this effect is heating the substrate directly is more effective than heating the substrate through the melted powder layer, and having both the substrate and molten powder at closer temperatures improves welding.

The partially exposed substrate effect links the roughness and bonding of the first layer and is less dependent on the laser parameters. For this reason, an alternate system was configured, such that multiple layers could be printed at each site, while still varying the fluence of the lasers from site to site. This was accomplished by motorizing the final turning mirror after lens(8), as shown in Fig. 1(a), to enable a multi-region imaging system similar in application to a recent Large Area Projection Micro Stereolithography (LAPµSL) technique [23]. Three layers at 20 different combinations of short and long pulse laser fluence were printed as shown in Fig. 4. An attempt was made to create the conditions shown at the right of Fig. 3 on the first layer to ensure good bonding of this layer. High resolution confocal images were recorded for all sites and the roughness was evaluated. The roughness values reported are the standard deviation of the heights located in a 200 µm square centered on each site (i.e., Ra roughness, see ISO 25178 standard) The square region of interest was chosen to avoid edge effects where the solid powder transitions down into the consolidated melted region, while still being large enough to accurately reflect mm-sized layers. With most powder particles being smaller than the 45 µm mesh they were sieved through, 200 µm is at least four melted powder particles wide and should account for surface features that change due to particle-to-particle relaxation during melting. In larger area prints, surface tension-driven effects may cause surface roughness which is not accounted for in this parameter sweep.

 

Fig. 4 Montage of optical micrographs showing 500 µm-wide pillars following the deposition of three layers. The surface roughness (Ra) as a function of long and short pulse fluence is indicated as insets. Only pillars that were confirmed to have 3 complete welded layers after removing the powder are labeled with roughness measurements. Optimal operating conditions for full builds is highlighted in the Fig. (2.1, 30.7 J/cm²).

Download Full Size | PDF

As either fluence is increased, the melted region becomes smoother, with hills and valleys from individual particles becoming less distinct (Fig. 4). At the lower short pulse fluences (e.g., 1.7 J/cm²) as the long pulse fluence is increased a threshold (~30.7 J/cm²) is reached where gaps appear in the third layer and part of the second layer can be seen through the gap. The formation of deposition gaps could be an indication of boiling or local fluid instabilities. At the highest short pulse fluence, no gaps are apparent at any long pulse fluences, indicating high short pulse fluence is favorable for reduced roughness. The roughness values are indicated on Fig. 4 in the lower right corner of each cell for all the sites that were confirmed to have three bonded layers after brushing away the powder with a soft plastic fiber brush. The lowest roughness occurred at 2.1 J/cm² and 30.7 J/cm² for the short and long pulse fluences respectively. Figure 5 shows the morphology of this site, with the top half showing an optical micrograph and the bottom half showing a confocal height map. Only a few solid particles remain and they may have been deposited on the melt pool as opposed to originating in the illuminated region and not melting. The exposed region is dominated by surface features with a spatial frequency much lower than the solid powder. The edge of the melted region is characterized by a sharp transition of only a couple of powder particles and no gap in the powder is apparent in the solid powder region outside the site.

 

Fig. 5 Split confocal image with optical micrograph on top and height map on bottom. A 30.7 J/cm² long pulse and 2.1 J/cm² short pulse were used to produce the deposition shown, and corresponds to the lowest roughness from Fig. 4.

Download Full Size | PDF

5. 3D Printing of metal components using DiAM

For demonstration of complete builds we used tin powder. Stainless steel 316L powder was also melted using fixed aperture masks, but is not reported here. The various inputs to the printing system can be split into two categories. Those that are specified before printing and those that change during printing. The first can be manually entered and the second must be scripted in time. The laser fluences were chosen based on the roughness measurements. The end of the short pulse laser was always delayed by 20 ms, so as to coincide with the end of the long pulse. The OALV was operated at 500 Hz and 30 V. These parameters were fixed throughout the printing process. The main scripted parameter was the pattern mask applied to the OALV between each layer.

Printed parts larger than the maximum single shot area of 2–3 mm could be achieved using stitching of images as mentioned above. Although ~10 mm wide hybrid laser beams could be imaged onto the OALV, the maximum practical area is limited by the short pulse laser which produces a non-uniform Gaussian intensity profile with only the central 2-3 mm uniform enough for our application. It should be noted however, that larger prints do not appear to be precluded by any bound in the physics of the process; both the laser beam and optics can easily be scaled to length scales approaching ~1 m. In order to print larger parts, an algorithm was used to generate OALV sub-masks that result in 2 × 2 mm² printed areas at the powder plane. The final mirror in the imaging relay optics between the OALV and powder plane was motorized, so that the sub-masks could be stitched together with 100 μm overlap. The position of the image at the powder plane is the second important parameter scripted in time.

The control software for the entire printing process is written in LabVIEW. The process begins with a 3D model in STL (Stereo-lithography) format, which contains the vertices for a triangular mesh representing the boundary of the region that should be solid in the printed part. The model is read from file and can be translated, scaled, or rotated on a virtual substrate. A view from the software is shown in Fig. 6, with the virtual substrate colored green and the model colored red. For each layer starting from the substrate and continuing at a specified interval (nominally 50 μm or the mean diameter of the powder) to the maximum height of the model, a plane parallel to the substrate is tested for intersection with every triangle in the model. For each triangle that intersects the plane, the resulting line segments are stored. The line segments are then sorted into a list, so that line segments with a matching vertex are adjacent. After removing the duplicate vertices, a list of vertices representing a polygonal line that defines a boundary between processed and unprocessed material is stored.

 

Fig. 6 3D rendering of an impeller part, being sliced and sub-sectioned for printing. Green indicates the substrate plate. Red is the impeller geometry. Blue indicates the area that will be illuminated. White lines outline the individual sub-sections that are stitched together.

Download Full Size | PDF

The projector that illuminates the OALV with the sub-section mask has pixels that after the image relay magnification illuminate an 8 × 8 μm area in the powder plane. Hence a virtual grid with 8 μm pixels is overlaid upon the contour lines generated from intersecting a virtual plane with the model. For each of pixels, a line extending across the entire row is intersected with every line segment in all contours. Then starting from one side of the row, all pixels are turned off until the first intersection, after which all pixels are turned on until the next intersection. This process repeats turning pixels on and off at each intersection until the end of the row. Every row is processed using this algorithm and the final output is a mask representing the area of powder that needs to be welded together at 8 μm resolution in the powder plane. The mask is then overlaid with the sub-section grid and individual regions of the mask are saved into an image to be loaded into the projector. An example mask and sub-section grid that was generated by the software is shown in blue and white in Fig. 6.

The motorized final mirror is out of an image plane in the optical system. We also note that the mirror is not tilted from the point where the imaging axis reflects from the mirror. This causes translation of the image at the powder plane in non-orthogonal directions and result in image rotation. These had to be accounted for in the control system. A deep bed of tin powder was exposed to a single layer of sub-sections, using only the long pulse laser at a fluence of 20 J/cm² that causes sintering but not melting. Where the powder is exposed multiple times the powder is further sintered and this is readily visualized by its reflectivity change in a visible light image. Figure 7 shows an example of a single layer stitching process after the mirror scanning and image rotation has been optimized.

 

Fig. 7 Wide area optical image captured during demonstration of the stitching control software. Long pulse diode light is incident on a deep powder bed, which enhances the contrast in the overlap region. The bright area corresponds to incandescence due to laser heating of the powder and molten material.

Download Full Size | PDF

We found that advancing the supply piston 10 μm, then moving the powder with the recoater at 100 mm/s to just in front of the substrate, followed by returning the recoater and repeating this process, then finally spreading the full dose of powder slowly over the substrate at 1 mm/s increased the layer uniformity and reduced drag marks. The build piston is moved down 50 μm to set the layer thickness before recoating and at the end of printing is lifted up to remove the substrate with the printed part welded to it. The entire spreading mechanism is placed in a standard glove box that is purged with argon resulting in an oxygen content <500 ppm by weight.

This new entire-layer-at-once 3D printing system was tested with two 3D models. The first is an impeller downloaded from Thingiverse.com and scaled down to fit our build envelope. The second was a CAD designed LLNL logo. Both are shown after printing in Fig. 8. The impeller model required 90 layers of printing with an average of 49 subsections per layer, requiring approximately 3.2 hours to build. The total illumination time with both lasers on, was less than 2 minutes, since the lasers are only printing for 20 ms per subsection. Ideally, stitching would be replaced by increasing laser power and image size so that each layer could be printed at once. This would reduce the total laser on time to 1.8 s. The remaining time for the print consisted of moving the image between sub-sections using mechanical motors (~2 s per layer) and operating the recoater (~30 s per layer). In comparison, for a 50 μm 1/e² diameter raster-scanned laser beam moving at 500 mm/s (typical in SLM systems), the total time for the build would be approximately 1 hour, and the total illumination time would be approximately 6 min. Thus, the efficiency in print time associated with use of the hybrid laser source is demonstrated here to increase by 3 × , with a potential to increase by 200 × using appropriate magnification optics.

 

Fig. 8 First demonstration of DiAM wide area photolithographic printing of metal layers using an optically-addressable light valve. For each build (impeller and LLNL logo) successive layers were built using a stitching approach which allowed additional efficiency to be achieved. Alternatively, scaling up overall dimension can be achieved by simply adding diodes and expanding optics, up to the energy limit of the Q-switched system (2 J).

Download Full Size | PDF

6. Conclusions

We have invented and demonstrated an alternative method for creating metal additively manufactured parts using a wide area, optically-addressable photomask supplied with a hybrid light source consisting of a low cost diode laser array and a Q-switched pulsed laser. The novel use of an optically addressed light valve not only allows extreme scaling and addressing on and off regions of the build but also offers the potential to ‘grayscale’ an image providing selective thermal gradients to be imposed potentially controlling residual stress and microstructure. Thus, the degree of design flexibility afforded through DiAM is far beyond that of typical selective laser melting systems and unparalleled in the overall metal AM market today. Surface roughness studies were conducted using DiAM to establish an optimal combination of Q-switched vs diode irradiation for fully bonded metal layers. A 13 mm test build was produced using a spatially-stitched laser beam, thereby demonstrating the efficacy of the technique. While the work reported here focused on a basic demonstration, commercial implementation of the technique for metal, polymer and/or ceramic materials has the potential to revolutionize metal additive manufacturing.