1 Introduction

Additive manufacturing (AM) has attracted considerable attention owing to its ability to produce complex parts with customised shapes [1]. Recent advancements in laser-based metal powder bed fusion (PBF-LB/M) have enabled the fabrication of sub-millimetre lattice structures, thereby expanding its application range [2, 3]. PBF-LB/M sub-millimetre structures (SMSs) are particularly interesting in biomedical applications, wherein the customisation of sub-millimetre-scale parts is becoming increasingly important [4, 5]. The most common SMS applications in the biomedical field include the fabrication of general custom-made implants such as bone-like scaffolds that promote bone ingrowth [6, 7] and the fabrication of small, patient-specific instruments [8]. Another highly promising and active research area in the biomedical field is the fabrication of cardiovascular stents through PBF-LB/M [4, 9,10,11].

Cardiovascular stents are conventionally fabricated through laser cutting using materials such as stainless steel, cobalt–chrome and nitinol. These stents are typically based on a honeycomb structure with strut thicknesses of 60–150 μm [9, 12]. However, conventionally manufactured stents suffer from limited geometrical flexibility because the stent designs are linked to a tubular precursor with a fixed diameter and wall thickness. The geometric mismatch between a straight stent and curved vessel anatomy often leads to local stress concentrations in the vessel wall, increasing the risk of late thrombosis [13, 14]. Nevertheless, patient-specific stents offer a precise fit of the stent design to the vessel anatomy, thereby resulting in a more uniform stress distribution and improved blood flow patterns that ultimately enhance the treatment outcomes [15, 16]. Using PBF-LB/M offers opportunities to produce patient-specific stents. Numerous studies have successfully fabricated 316 L stainless steel, cobalt–chrome and nitinol stents that meet high-quality standards through conventional PBF-LB/M using commercially available machines [10, 11, 16]. Compared with micro PBF-LB/M, conventional PBF-LB/M offers several advantages, such as higher process stability, availability and cost-effectiveness, endowing it with more potential for industrial use. Finazzi et al. developed a novel mesh design specifically optimised for PBF-LB/M-manufactured stents and demonstrated the expansion of a cobalt–chrome stent through balloon expansion [9].

Although PBF-LB/M has shown promise for the fabrication of patient-specific stents, it still encounters some fundamental challenges, namely dimensional process limitations, poor surface quality, difficulty in removing support structures and low geometrical accuracy. The minimal achievable printing size of 200–300 μm in PBF-LB/M is restricted by laser spot size (30–100 μm), powder particle size (10–50 μm) and layer height (20–50 μm) [11]. Recent studies have attempted to overcome these limitations by focusing on micro-part printing using μPBF-LB/M [17]. However, micro-printed stents may exhibit poor mechanical and functional properties and fine powder usage poses a further safety risk [11, 18].

PBF-LB/M parts often exhibit high surface roughness owing to the layer-by-layer nature of the process and the thermodynamic phenomena occurring during material deposition and fusion [19, 20]. A low surface roughness (Ra < 0.5 μm) is important in achieving adequate hemocompatibility in biomedical applications since it can negatively influence the mechanical performance of stents [21].

Support structures are necessary for connecting the stent to the base plate, promoting heat dissipation [22] and stabilising the filigree part during printing [23]. However, the mechanical removal of supports can deform or damage the filigree stent structure and further aggravate the down-skin surface roughness [24]. Therefore, self-supporting structures that avoid the usage of supports altogether must be designed [25]. However, this limits the fabrication of open-cell structures, which are preferred for stents. Lack of geometrical accuracy can substantially affect the mechanical or functional properties of stents [5, 21]. The process-related effects associated with geometrical deviations (GDs) include the staircase effect and the presence of partially sintered particles at low strut inclination angles [21, 26, 27]. Coater-dependent effects, such as scrape-induced thinning and compression-induced thickening, affect the printed strut thickness [26]. Recent studies have addressed this issue by implementing computational prediction and compensation models for lattice structures or applying electrochemical post-processing to compensate for GDs [21, 26, 28].

Some of the aforementioned challenges have been addressed by improving the print quality of PBF-LB/M-fabricated stents using optimised process parameters and establishing specific design rules for AM stents [11, 25]. However, these issues cannot be completely solved by improving the design and manufacturing processes alone. Many studies that have fabricated PBF-LB/M stents used the electrochemical post-processing methods of electropolishing or chemical etching to reduce printed strut thickness, improve surface quality and remove supports within the stent [5, 9,10,11, 16]. Although these studies have demonstrated the potential of conventional electrochemical post-processing methods to address the challenges associated with producing filigree SMSs using PBF-LB/M, the results remain unsatisfactory. A more systematic evaluation of the printing process and the customisation of post-processing parameters to the specific requirements of PBF-LB/M-fabricated filigree SMSs are clearly needed. Furthermore, the effect of the design aspects on the print quality and material removal interactions needs further investigation.

The use of hybrid post-processing treatments for PBF-LB/M-fabricated parts is attracting increasing research interest [20, 29, 30]. Using these hybrid treatments allows combining the advantages of different post-processing methods in a single fabrication step, thereby providing improved results in less manufacturing time. Hirtisation® is an (electro)chemical hybrid post-processing technique (HPPT) developed by RENA Technologies (Austria). This technique is specifically designed for the post-processing of AM metal parts [31,32,33,34] through dynamic electrochemically and hydrodynamically assisted chemical reactions to precisely control material removal and surface smoothing in two subsequent treatment steps [35]. Although Hirtisation® is primarily applied to large and complex hollow parts, its potential for processing filigree SMSs remains largely unexplored [34].

This study mainly aims to investigate the potential and limitations of Hirtisation® as an HPPT to overcome the challenges associated with the PBF-LB/M manufacturing of filigree SMS. First, this study specifically focuses on the HPPT process considerations by comprehensively investigating the HPPT process parameters to identify suitable parameters for processing filigree SMSs. Next, design considerations are addressed by examining the effect of various design aspects on material removal through the HPPT and the resulting geometric accuracy of the processed parts. Accordingly, two PBF-LB/M sample designs are conceptualised, manufactured, post-processed and evaluated. Finally, the important key considerations for manufacturing high-quality and accurate filigree SMSs are highlighted.

2 Methods

2.1 Test samples and procedures

This study comprised two consecutive parts and a formulation of the key considerations required when applying the HPPT to filigree SMSs (Fig. 1). In the first part, process considerations were made to determine the suitable process parameters for post-processing of filigree SMSs (e.g. vascular stents). Using an application-oriented approach, the stent samples were designed to achieve a fixed geometry. Through an iterative process, variable HPPT process parameters were applied to these samples to enhance the surface quality, facilitate the damage-free removal of support structures and overcome the dimensional process limitations.

Fig. 1
figure 1

Overview of the consecutive study workflow. The first part comprises the design considerations made to define the adequate HPPT process parameters for filigree SMSs. These parameters are transferred to the second part to investigate the influence of different design aspects on material removal through the HPPT. Finally, the key considerations for the HPPT in case of filigree SMSs are formulated

Full size image

The second part focused on design considerations, whereby the identified HPPT process parameters were applied to a representative test geometry for filigree SMSs. Generic test samples with varying design aspects were derived from the stent sample to systematically examine the potential and limitations of the HPPT in mitigating the process- and design-induced GDs to gain a comprehensive understanding of the material removal mechanism through the HPPT, which is considerably affected by various factors (e.g. initial printing quality, specific machine, material and printing parameters) and design aspects (e.g. strut build orientation, strut inclination angle, strut thickness and support implementation along the strut). The approach used for the process and design considerations will be described in more detail in the subsequent sections.

2.1.1 Process considerations

The stent sample designs for the first part of the study were based on a simple honeycomb structure considering the design rules of Finazzi et al. [25]. These rules were originally developed to fabricate expandable stent meshes and included specifications, such as a 1 mm maximum overhang length, a 45° maximum strut overhang angle and a 0.3 mm minimum spacing between struts.

A 10 mm stent diameter was selected to facilitate handling and measurements. The samples were printed on a round base, allowing them to be connected to and removed from the build plate without damage. The element struts were inclined at a 55° angle. The strut length was limited to 4 mm, resulting in a ring of six elements per level. These rings were arranged in two configurations, i.e. aligned peak to valley and peak to peak (Fig. 2a). Every other vertical link between the nodes of two elements was designed as a single support structure to be fully removed via post-processing. The supports were designed with a 300 μm minimum achievable printed thickness, for which a 200 μm minimum material removal was needed to remove the supports.

Fig. 2
figure 2

a CAD model of the stent samples and visualization of the surface roughness measurement strategy according to DIN EN ISO 3,274:1998-04 and the strut thickness measurement ‘t’. b As-printed samples. ‘SR’ is used to investigate the support removal (S) and the surface quality (R). ‘T’ is utilised to investigate the minimal achievable strut thickness (T)

Full size image

The nominal strut thickness (tCAD) was calculated by considering the desired post-processed strut thickness (tPP) and the necessary support removal thickness (tSupRem) as follows:

$${t}_{\textrm{CAD}}={t}_{PP}+{t}_{\textrm{SupRem}}.$$
(1.1)

Considering that the printed thickness results larger than the designed thickness, the to-be-expected as-printed strut thickness (tAP) is defined as follows:

$${t}_{\textrm{AP}}\ge {t}_{PP}+{t}_{\textrm{SupRem}}.$$
(1.2)

The final strut thickness was further fine-tuned by increasing the processing time and removing more materials. Fine-tuning (tFinetune) was also needed to remove the residuals left from the supports and reduce the surface roughness. Accordingly, the to-be-expected post-processes strut thickness (tPP) is defined as follows:

$${t}_{\textrm{PP}}\le {t}_{\textrm{AP}}-{t}_{\textrm{SupRem}}-{t}_{\textrm{Finetune}}$$
(1.3)

This calculation approach allowed for a precise control over the final strut thickness. Two different sample types were fabricated. Surface roughness (SR) samples were used to test the support removal and the surface quality. The strut thickness was defined at 1 mm to account for the unpredicted occurrences during post-processing that might require large material removal. The thickness (T) samples were designed with a smaller strut thickness of 500 μm to test the minimum achievable strut thickness through the HPPT. Five samples were printed for each type, with four post-processed and one used as the as-printed reference. Figure 2b illustrates the as-printed condition of the two different sample types.

2.1.2 Design considerations

Test samples with angled struts were designed to represent the main stent features for systematically investigating the influence of certain design aspects on print quality and geometric accuracy (Fig. 3b). The investigated design aspects are strut thickness, strut overhang angle, strut build orientation, and integrated supports. The flat design of the test samples was selected to minimise the measurement errors associated with a curved surface. The struts were built on a rectangular base that served as a contact interface for post-processing and a solid connection to the build plate that enabled damage-free sample removal. Three main struts on each side of the sample were designed to investigate the effect of the strut build orientation, with the left side struts tilted against (TA) and the right side tilted with (TW) the coater direction. A vertical strut (0°) was incorporated in the middle as a reference, while a horizontal strut (90°) was placed at the top of the sample to test the bridging potential of the PFBF-LB/M process. Three sample types with different strut inclination angles of 45°, 60° and 75° were designed to examine the effect of the strut inclination angle. Three support strategies were generated to test whether the supports can minimise the printing-induced GDs. Figure 3a) provides an overview of the tested sample clusters. Four versions were tested for each of the three basic angle designs: one without supports (NS), one with supports spaced 0.3 mm apart (S03), one with supports spaced 0.9 mm apart (S09) and one with a 0.3 mm spacing between the supports and a four-layer gap between the support and the strut (S03G). The exposed struts at the sample edges allowed the assessment of the surface quality at the top and bottom sides of the struts after support removal. All presented clusters were printed with two different thicknesses (i.e. 0.6 mm, which was the smallest tCAD possible to use (Eq. 1.3) and 0.9 mm, which was a reliable printing thickness on the used machine) to address the strut thickness effect. The number of samples tested for each cluster was n = 3.

Fig. 3
figure 3

Overview and CAD model of the test samples. a Overview of the tested clusters depending on the support strategy and the thickness: NS, non-supported; S03, 0.3 mm spacing; S09, 0.9 mm spacing; and S03G, 0.3 mm spacing with a four-layer gap. These groups are tested for the three angles of 45°, 60° and 75° and for the two strut thicknesses of 0.6 and 0.9 mm. b CAD of a supported 45° sample containing struts tilted with (TW) and tilted against (TA) the coater direction. tSupport was always constant at 0.3 mm. Spacing is the distance between the outer edge of the two supports. Gaps were inserted between the top and the bottom of the support and the strut to be supported

Full size image

2.2 Sample material

CL 92PH precipitation hardening stainless steel from GE (Boston, US) was used for all study samples. This steel alloy powder form is known as 17-4PH and has a chemical composition according to presented in ASTM A564/A564M – 13 UNS S17400/SUS 630. Table 1 presents the chemical composition of the powder.

Table 1 Nominal chemical composition of the 17-4 PH powder expressed in weight percentage (wt.%)
Full size table

2.3 Laser-based metal powder bed fusion

All test samples were manufactured using a Concept Laser M2 Cusing R machine (Lichtenfels, Germany) equipped with a Yb:YAG laser with a 1,070 nm wavelength. The laser provided 100 W of maximum power and had a focus diameter of 50 μm. The machine print volume was 90 mm × 90 mm × 80 mm. The layer height used in all prints was 30 μm. The system operated in a protective argon atmosphere. The samples were positioned and sliced with Magics 19 from Materialise (Leuven, Belgium). At the core, an island laser scanning strategy was applied surrounded by a concentric, approximately 120 μm-broad shell line. The stent samples were placed with the label perpendicular to the coater direction, whereas the test samples were placed parallel to the coater direction with a 3° offset (Fig. 3b).

2.4 Electrochemical hybrid post-processing: Hirtisation®

2.4.1 Choice of post-processing method

Conventional electrochemical post-processing methods such as electropolishing or etching often encounter challenges related to surface quality, precision, and process efficiency [24, 32]. These challenges are particularly pronounced for filigree SMS, such as stents [5, 9].

Several studies have explored electrochemical post-processing techniques, such as electropolishing or chemical etching, in this regard [5, 9,10,11, 16]. These investigations aimed to reduce the printed strut thickness, improve surface quality, and remove supports within the stent. While these studies have demonstrated the potential of conventional electrochemical post-processing methods in addressing these aspects, the results remain unsatisfactory.

Hirtisation® is an advanced hybrid post-processing technique developed by RENA Technologies (Wr. Neustadt, Austria). In contrast to conventional methods, it combines two different electrochemical processes in subsequent treatment steps [31, 32]. Specifically, the method leverages dynamic electrochemically and hydrodynamically assisted chemical reactions to precisely control material removal and surface smoothing respectively. The hybrid nature of a HPPT makes it possible to integrate the benefits of different post-treatment processes into a streamlined process. Consequently, Hirtisation® can substantially improve surface quality, reduce production time, and increase overall process efficiency [32, 33]. The fundamental principle of the novel HPPT method and how it leverages these process-related advantages are described in Section 2.4.2.

2.4.2 Principle of Hirtisation®

Chemical and electrochemical methods involving liquid electrolytes play a crucial role in enhancing surface quality. These methods rely on carefully controlled parameters, such as concentration, solubility, reaction rate, material load, throwing power and reachability [36]. Based on the selected parameter, electrolytic treatment can offer various functions such as descaling, pickling, chemical etching, deburring, electrochemical machining and electropolishing; further, electropolishing is widely used for surface improvement [37, 38].

The extent of the process reactivity is predominantly determined by physical parameters such as concentration, temperature and exposure time. Secondary parameters such as potential, current density and pulsed signals can enhance efficiency and surface quality. Although liquid media are advantageous for treating complex and intricate surfaces, they have limitations. For instance, during electropolishing, the surface static potential varies, leading to non-uniform treatment. Transitioning from etching to polishing is challenging owing to the dominant role of electrolyte chemistry in the process. The electrochemical conditions on a substrate surface can be changed by manipulating the current and potential fields. Altering pulse parameters allows the process to change from etching to polishing, enabling precise removal of excessive structures while achieving a polishing effect. Hirtisation® addresses these challenges by providing a sequence of treatment stages tailored to the initial quality of AM parts and the desired final properties for their application.

2.4.3 Post-processing protocol

Before being further processed by the HPPT, all test samples underwent 5 min of ultrasonic cleaning for the removal of the loose powder particles. The HPPT was performed using an H3000s system from RENA Technologies (Wr. Neustadt, Austria).

Figure 4 depicts the HPPT setup. The test samples were immersed in an electrolyte solution and connected to a positive power supply (anode) terminal using either a clip (stent samples) or a screwed (test samples) connection. The counter electrode (cathode) was submerged in the electrolyte solution at the sides of the bath. The stent samples were individually processed and positioned in the middle of the bath, while the test samples were distributed over one row containing seven to eight samples for each batch (Fig. 4). The test samples were shielded with plastic tubes to prevent them from severe material removal at the sample tip and to achieve a more homogenous material removal along the part.

Fig. 4
figure 4

Schematic representation of the HPPT and the corresponding components for the test sample processing. The stent samples were individually processed and did not include the shield at the sample tip

Full size image

Due to the overall number of samples exceeding the capacity of the Hirtisation chamber, it was necessary to split the samples into smaller batches. The distribution of samples across different batches is shown in Table 2. The samples were distributed such that batch-related effects could be minimised and evenly distributed among the sample groups. Each batch contained three complete sample clusters of five samples with the same strut thickness and support strategy, but different angles (corresponding to a row in Fig. 3). This approach ensured that batch-related effects, if present, would impact entire sample clusters rather than individual samples, thereby reducing the risk of outliers within a cluster and improving comparability among the sample clusters.

Table 2 Sample grouping in the different HPPT batches of the test samples for the second part. All batches were executed with the same process parameters. The number of samples of the two batches was n = 15
Full size table

An overview of the applied processing parameters is shown in Table 3. The mechanism of material removal, as well as the iterative definition of temperature and time used for the HPPT, will be explained in the following.

Table 3 Applied process parameters of the HPPT for steps I and II
Full size table

The HPPT involves two distinct steps utilising different electrolyte solutions based on an acidic medium and distinct current distributions to achieve precise material removal and surface smoothing (Appendix 1). The use of different current distributions allows control over angle- and area of attack (Fig. 5) within the process and enables adjustment between more aggressive and more precise material removal [35].

Fig. 5
figure 5

Overview of the three current distributions and their process effects, with δ being the diffusion layer thickness. Adopted from [39]

Full size image

Macroscopic material removal (step I): The first step cleaned the passive surface layers and removed the macroscopic excessive structures by partially exploiting the primary current distribution. This step effectively targets and reduces larger surface irregularities and partially sintered particles. Furthermore, during this step, the strut thickness was reduced to a large extent. This step ensures efficient and broad material removal, thereby decreasing the overall surface roughness to a manageable level.

Surface polishing (step II): Once the surface roughness was reduced to a certain level, it became possible to proceed to the polishing step with a different electrolyte and voltage control. Homogenous polishing was achieved using a combined secondary–tertiary current distribution field owing to the complex and combined bulk and filigree geometry of the samples. Hence, the original geometry and shape of the sample were preserved.

In addition to the electrolyte composition and the current distribution adjusted by the applied current pulses, the main influencing parameters shaping the process are temperature and time. The processing temperature was set at 70 °C and 60 °C for steps I and II, respectively. The processing times are highly dependent on the part size and geometry; thus, an iterative approach was used to determine the ideal duration of the two process steps to achieve a residual-free support removal and the best possible surface quality. The other parameters remained constant. First, the adjustment parts were processed using step I until the supports were removed and the surface roughness was close to Ra = 2 μm (Fig. 6). Step II was applied until smoothing reached a plateau. This resulted in initial processing times of 60 and 10 min for steps I and II, respectively.

Fig. 6
figure 6

Overview of all post-processed stent samples: top: surface roughness (SR); bottom: strut thickness (T). The table features the corresponding processing times of both steps in minutes, the average as-printed and post-processed Ra values and the average material removal and post-process strut thickness in millimetres

Full size image

The working set for the test samples in the design consideration part was defined based on the stent sample results (Fig. 6). The best results were achieved with processing times of 70 and ≥10 min for steps I and II, respectively. The adjustment parts were processed beforehand, and the processing times were defined at 60 and 25 min for steps I and II, respectively, to adjust the parameter set to the increased number of supports and achieve good results in the case of all the support strategies.

2.5 Print quality measurements

The longitudinal internal porosity was investigated to determine the bulk print quality of the samples, as it can highly affect post-processing outcomes. The porosity samples, which contained lamellae printed on a rectangular base representing the struts, were designed with thicknesses of 0.3, 0.5 and 1 mm. To additionally evaluate the effect of the inclination angle on the bulk print quality, the samples were printed via PBF-LB/M at inclination angles of 45°, 55° and 90°. The 90° angle, with no overhang, was chosen as a reference, while the 45° angle was selected as the most extreme permissible overhang [1, 25]. The 55° angle was included to assess the effect of a more conservative angle on print quality, especially at very low strut thicknesses. Subsequently, the samples were encapsulated in resin, ground, and polished to obtain a highly reflective surface (Struers, Germany). The polished cross-sections were observed via digital microscopy (VHX-6000, Keyence, Japan). The sample porosity manifested as dark voids on the cross-sectional surfaces (Fig. 8).

2.6 Laser profilometry

The surface roughness measurements before and after the HPPT were acquired using an optical surface measurement device (Infinite Focus SL 3D step, Alicona, Austria). The surface roughness profiles were captured using a three-dimensional laser scanning microscope (VK-X100, Keyence, Japan). According to DIN EN ISO 3,274:1998-04, the present Ra range of the as-printed and post-processed test samples between 2 and 10 μm required a total measurement length of 12.5 mm. A multi-line roughness profile measurement was performed considering the limited strut lengths and resulted in five 2.5 mm parallel measurement lines stitched together (Fig. 2a).

The surface roughness was measured at the base of the stent samples. The surface curvature was compensated for in the associated Keyence software to avoid distortion effects and a possible negative impact on the surface roughness measurement.

2.7 Digital microscopy

Images for assessing the geometrical accuracy of the as-printed and post-processed samples were obtained using digital microscopy (VHX-6000, Keyence, Japan). The stent samples were oriented such that the struts were captured parallel to the baseplate, resulting in a 150× magnification for the vertical struts and 100× magnification for the angled ones. The test samples were placed flat under a microscope and captured at 150× magnification. Stitching of multiple images and depth composition was enabled to overcome the limited field of view resulting from the high magnification and blurring at the edges.

The thickness measurements were manually performed using Keyence software by placing two parallel lines on either side of the measured strut. To achieve reproducible measurements, clear references were defined on the surface topology from which the measurement lines could be established. For the as-printed struts, the deepest valleys of the outermost outline were selected as the reference. For the post-processed struts, the middle line between the peaks and valleys of the outermost outline was selected.

3 Results and discussion

3.1 Process considerations

Figure 6 illustrates a comprehensive overview of the iteratively processed samples and their corresponding measurement results. The as-printed samples exhibited a homogenously printed surface topology and an average surface roughness of Ra = 5.1 ± 0.8 μm for both the SR and T samples. However, small artefacts were also observed at the overhanging areas near the node edges, where complete overhang printing occurred. The results presented in Fig. 6 indicate that under all conditions, the supports could be successfully removed and the surface roughness decreased after HPPT processing. The findings for the print accuracy, internal print quality, dimensional process limitations, support removal and surface roughness will be presented and discussed in more detail in the subsequent sections.

3.1.1 Coater-induced effect

Strut thickness inhomogeneities were observed along the whole samples. The angled struts specifically deviated more from the nominal thickness when compared with the vertical references. A trend was also identified, wherein the printed strut thickness (tStrut) varied based on the build orientation relative to the coater direction of the angled struts owned to the circular stent design (Fig. 7). The plots of strut thickness as a function of strut number presented in Fig. 7b show that all TA struts (i.e. 2, 4, 5 and 7) exhibited peaks, whereas the TW struts (i.e. 3, 6 and 9) exhibited valleys. The struts oriented parallel to the coater direction (i.e. 1 and 8) exhibited the lowest and highest minima and maxima. The observed deviations were attributed to the different process-related effects. The resulting wider thicknesses of the angled struts compared with those of the vertical struts were attributed to the staircase effect, the partly sintered particles attached to the down skin surfaces and the occurring thermal effects [27, 40]. The fluctuations that depended on the strut orientation were caused by the additional coater-dependent effects, including compression-induced thickening and scrape-induced thinning, which led to thicker and thinner struts depending on the strut orientation relative to the coater direction [26]. Figure 7 Strut thickness (tStrut) evolution of the as-printed (AP) and post-processed (PP) stent samples dependent on the orientation of the angled struts relative to the coater direction. a Overview of the strut numbers and the inclination direction (from bottom to top) of the strut indicated with red arrows. b Quantitative representation of the measured strut thicknesses of the surface roughness (SR) and minimal thickness (T) samples (Fig. 2b) View.

Fig. 7
figure 7

Strut thickness (tStrut) evolution of the as-printed (AP) and post-processed (PP) stent samples dependent on the orientation of the angled struts relative to the coater direction. a Overview of the strut numbers and the inclination direction (from bottom to top) of the strut indicated with red arrows. b Quantitative representation of the measured strut thicknesses of the surface roughness (SR) and minimal thickness (T) samples (Fig. 2b)

Full size image

3.1.2 Bulk print quality

The porosity samples were analysed to evaluate the thickness and angle-dependent bulk print quality. Results show that the void content of the cross-sections was considerably low, with an average relative density of > 99.1% observed for all thicknesses and angles. However, a void zone was observed near the strut surface, located at approximately 115 ± 10 μm beneath the surface, containing voids that reached up to a 50 μm diameter (Fig. 8). This void zone can be attributed to the scanning strategy used because the transition between the shell and core trajectories is assumed at approximately 120 μm from the outer strut edge. Since the results indicate that different inclination angles did not show significant differences in print quality, the minimum angle of 45° was implemented in the stent samples.

Fig. 8
figure 8

Digital microscopy images of the porosity samples with different thicknesses of 0.3, 0.5 and 1 mm and print inclination angles of 90, 55 and 45°. Left: strut cross-sections. Right: magnification of the marked areas in the 90° and 55° struts

Full size image

3.1.3 Dimensional process limitations

Figure 6 displays the detailed processing times and thicknesses of the samples. An average strut thickness of 150 ± 0.03 μm was achieved (T3) with 70 and 10 min of processing times for steps I and II, respectively. An alternative approach with a reduced processing time of 40 min for step I and a prolonged processing time of 20 min for step II resulted in stent meshes with an even lower average strut thickness of 110 ± 0.04 μm (T4). Furthermore, the SR4 and T4 samples exhibited non-homogenous strut thicknesses along the part with thinning occurrence at the sample tips Fig. 9, indicating that the prolongation of step II led to inhomogeneous material removal. Hence, an approach with a higher processing time for step I instead of step II should be favoured. The HPPT parameters must be carefully selected because a slight difference can lead to sample detachment from the base (T2) or mesh distortion (T4).

Fig. 9
figure 9

Digital microscopy images of the as-printed and post-processed SR and T samples and a qualitative visualisation of the corresponding processing times for both HPPT steps. a SR samples showing a successful support removal and the effect of the remaining support residuals and thinning at the sample tips. b T samples showing a successful support removal, strut thinning to 110 μm and the effect of the remaining support residuals and thinning at the sample tips

Full size image

The inhomogeneous material removal at the top struts was attributed to the inhomogeneous distribution of the current field lines under the secondary current regime. Prevention of this phenomenon was attempted by implementing shields for test sample processing in the second part of the study (Fig. 4).

The achieved homogenous strut thickness of 150 ± 0.03 μm was smaller than the strut thicknesses reported to date. Demir et al. achieved an average strut diameter of 250–350 μm through improved printing by investigating different scanning strategies [11]. Similar to the present study, Orla et al. achieved an average strut thickness of 152 μm through post-processing with chemical etching [16].

3.1.4 Support removal

The digital microscopy images of the post-processed SR and T samples presented in Fig. 9 show that the supports were successfully removed using the HPPT, i.e. open-cell designs could be generated, and damage-free part removal from the base was enabled. However, the support removal quality is influenced by the process parameters. As depicted by the bars in Fig. 9, the longer processing times for step I resulted in fewer residuals at the nodes. An increase in the processing time for step II also led to lower surface roughness and smoother node edges. However, the higher processing times for step II led to more inhomogeneous strut thicknesses (Section 3.1.3). Thus, the processing time cannot be excessively increased. Support removal was best achieved with processing times of 70 min for step I and a maximum of 20 min for step II. This parameter set resulted from a combination of the processing times of the SR4 and T3 samples shown in Fig. 6.

3.1.5 Surface roughness

Figure 10 shows the digital microscopy images and the corresponding roughness profiles of one as-printed and two post-processed SR samples. Analogous to the samples shown, all the samples exhibited homogeneous roughness profiles for both thicknesses. However, the resulting surface roughness was highly dependent on the applied HPPT process times. The graph presented in Fig. 10 shows the relationship among the cumulated process times, material removal and resulting surface roughness for the three samples. The surface roughness first increased and then decreased with increasing process times and material removal. This can also be seen in the roughness profiles presented in Fig. 10. A more detailed analysis of the material removal revealed that the process depth of the material removal for the samples with high surface roughness values between 5.8 and 18.8 μm (i.e. SR2, SR3 and T3) reached the identified porous layer at 120 μm depth (Section 3.1.2). This observation was also made in the case of the SR2 sample, whose roughness profile showed a crater-like pattern indicating that the process opened the void zone at the strut edges.

Fig. 10
figure 10

Quantitative representation of the relationship of the cumulated processing times and the corresponding surface roughness and material removal values of an as-printed SR1 and post-processed SR2 and SR4 samples. The digital microscopy images of the samples are shown below, with their corresponding surface morphologies. As-printed SR1 strut with a homogenous and rough surface morphology and Ra = 4.4 μm. Post-processed SR2 strut showing a crater-like pattern with Ra = 18.8 μm. Post-processed SR4 strut showing a smoothened surface with Ra = 2.0 μm

Full size image

Step I of the HPPT was primarily designed to remove the partially sintered particles and a substantial amount of material. Therefore, the processing time for step I must be carefully selected to prevent internal porous layer penetration. After reaching an ideal process depth, further material removal and additional smoothing can be achieved using step II. The SR4 and T4 samples that underwent the longest processing times for step II (i.e. 50 and 20 min, respectively) exhibited Ra values of ≤ 2.1 μm (Fig. 6). However, as discussed in Section 3.1.3, the longer processing times for step II resulted in an inhomogeneous strut thickness along the part. Therefore, the best achievable surface roughness was expected to be obtained with processing times of 70 min for step I and a maximum of 20 min for step II. These times were consistent with the times identified for the successful support removal in Section 3.1.4.

The lowest surface roughness value achieved herein was Ra = 1.9 μm for sample T3 that underwent the HPPT with the identified ideal processing times of 70 min for step I and 10 min for step II. This Ra value was at the lower range limit reported in previous studies. Demir et al. and Finazzi et al. reported on electrochemically polished SMS achieving roughness values between Ra = 1.45 ± 0.09 and 5.96 μm. Orla et al. achieved an average roughness value of only 0.36 ± 0.06 μm through chemical etching [9, 11, 16]. To the best of the authors’ knowledge, this is the only study that achieved the required surface roughness of Ra < 0.5 μm for medical implants using conventional PBF-LB/M to produce stents. Notably, the HPPT outcome is highly dependent on the initial print quality, particularly the internal porosity. Therefore, even better roughness values are expected with a higher print quality of the parts.

3.2 Design considerations

3.2.1 As-printed quality and accuracy

Figure 11 shows the effects of the strut inclination angle and support structures on the printing accuracy of the strut thickness. A comparison of the vertical 90° struts using the nominal CAD revealed a process-related fundamental average GD of 67 ± 6 μm thickness for the 0.6 mm samples and 71 ± 6 μm thickness for the 0.9 mm ones. A clear trend was also observed for both strut thicknesses, with the GDs increasing even more as the inclination angle of the strut decreased. Overall, across the support strategies, the struts with a 45° inclination angle exhibited an average GD of up to 189 μm for the 0.6mm samples and 200 μm for the 0.9 mm samples. For all the other angles, the 0.9-mm samples exhibited slightly higher GDs. The differences in the average GDs per angle between the 0.6 and 0.9 mm samples falling within the 4.22 (90°)–16.58 μm (45°) range indicated that the strut thickness did not substantially affect the GDs. The implementation of additional supports also did not significantly reduce the angle-dependent GDs.

Fig. 11
figure 11

As-printed strut thicknesses clustered by the strut inclination angle and the support strategy. a Results for the 0.6 mm-thick samples. b Results for the 0.9 mm-thick samples

Full size image

The observed fluctuations of the GD depending on the angle were likely caused by the staircase effect and the partially sintered particles adhered to the down skin surface of the struts [27]. The presence of long whiskers in the measurements of both sample thicknesses suggested a considerable variability. This variability may be attributed to the manual measurement method and other printing-induced effects, such as the strut orientation relative to the coater direction [26, 28].

Further analysis on the as-printed design accuracy was performed by dividing the boxes into TA and TW struts, which revealed a notable difference in the strut thickness depending on the strut orientation. Calculating the averages of all support strategies for each angle in the 0.6 mm samples (Fig. 12a) showed that the TA struts exhibited average strut thickness increases of 68.8 ± 24, 59.2 ± 17 and 30.0 ± 9 μm (45°, 60° and 75°, respectively) compared with the TW struts. Similarly, the TA struts in the 0.9 mm samples were 66.8 ± 34, 60.2 ± 16 and 27.5 ± 16 μm thicker compared to the TW struts.

Fig. 12
figure 12

As-printed strut thicknesses of the struts facing towards and away from the coater direction clustered by the strut inclination angle and the support strategy. a Results for the 0.6 mm-thick samples. b Results for the 0.9 mm-thick samples

Full size image

The smaller whiskers observed in Fig. 12 compared to Fig. 11, where all struts were plotted together in one box, indicated that the measurement variability decreased. In other words, most of the observed variabilities in Fig. 11 resulted from the pressure-induced thickening at the TA and the scrape-induced thinning at the TW struts [26]. The deteriorated alignment of the individual boxes at inclination angles of 45° and 60° for both the sample thicknesses demonstrated that the additional supports did not diminish the printing-induced GD. Conversely, the additional supports appeared to generate more strut thickness variabilities.

3.2.2 Material removal

Figure 13 reveals the findings on the effect of the strut inclination angle and the support strategy on the material removal through the HPPT for both sample thicknesses. The reported material removal was calculated as follows: tAs-printed - tPost-processed /2, with t as the strut thickness. A comparison of the average values of each bar between the 0.6- and 0.9-thick samples showed that the average material removal in the 0.9 mm-thick samples was up to 18.5 μm higher compared with the 0.6 mm-thick samples. Furthermore, the material removal for both sample thicknesses tended to decrease with increasing strut inclination angle, particularly beyond the 75° strut inclination. In contrast to the strut printing accuracy, the used support strategy seemed to influence the material removal. In the case of the 0.6 mm-thick samples, the least material removal occurred with the 0.3 mm spaced supports (S03). For the 0.9 mm-thick samples, the least material removal was observed with the 0.9 mm spaced supports (S09). The highest material removal was observed with the 0.3 mm supports having a four-layer gap (S03G) (0.6 mm samples) and without supports (0.9 mm samples). However, the 90° struts of both sample thicknesses showed some notably high fluctuations indicating the additional presence of batch-related differences in material removal. The batches of the 0.3 mm spaced supports with the four-layer gap (S03G) showed a higher material removal compared with the other sample batches. Considering these effects and normalising the absolute material removal by subtracting the basic material removal occurring at the 90° struts, the highest material removal for both sample thicknesses occurred for the samples without supports, while the least material removal occurred for the samples with supports that did not have an additional four-layer gap. In other words, a higher material removal was observed when fewer supports, larger spacing and an additional four-layer gap were implemented. Thicker struts and a higher strut-to-support ratio tended to increase the material removal as well. These results were attributed to the better electrolyte access through larger spacing and fewer supports as well as a better current flow to the supports through the thicker struts of the samples. Furthermore, the increased material removal for higher inclination angles was due to the decreasing surface quality at lower inclination angles. This decreased surface quality tended to exhibit more sintered particles and a larger surface area for the HPPT to act upon, consequently resulting in higher material removal.

Fig. 13
figure 13

Material removal calculated from the as-printed thickness subtracting the post-processed strut thickness clustered by the strut inclination angle and the support strategy. a Results for the 0.6 mm-thick samples. b Results for the 0.9 mm-thick samples

Full size image

Figure 14 provides further insights into the coater-dependent effect on material removal by examining the results based on the strut orientation. The trends observed in Fig. 13 remained consistent for both sample thicknesses. The material removal also appeared to be influenced by the strut orientation depending on the inclination angle. The difference in the material removal between the TA and TW struts diminished as the inclination angle increased, particularly for the 0.9 mm-thick samples where the average difference became <10 μm at a 75° inclination angle. At higher inclination angles, the TA struts tended to exhibit greater material removal than the TW struts for all support strategies and both sample thicknesses. However, this trend was reversed at a 45° inclination angle and material removal increased on the TW struts. In addition to the known coater-induced effects [26], the observed phenomenon can also be attributed to the powder distribution in the powder bed. For the TA struts, the powder may become lodged beneath the overhanging geometry and result in additional compression. This could increase the powder density within the struts and potentially enhance their resistance to material removal during the HPPT process. However, this hypothesis remains untested and requires further investigation through detailed simulations of the coating process during printing.

Fig. 14
figure 14

Material removal of the struts facing towards and away from the coater direction clustered by the strut inclination angle and the support strategy. a Results for the 0.6 mm-thick samples. b Results for the 0.9 mm-thick samples

Full size image

Appendix 2 presents a comparison of the material removal of all struts of the whole sample with that of the upper two struts of each sample. The average material removal without the upper struts was lower than that containing all the struts. Moreover, the average material removal without the upper struts mostly depicted reduced variability. Hence, despite the implementation of shielding measures to protect the upper struts from electrolyte exposure, the analysis revealed a location-dependent effect on the material removal through the HPPT. The results confirmed the findings presented in Section 3.13. indicating that the strut position considerably affects the material removal, irrespective of the sample thickness, inclination angles, and support strategies employed. Therefore, further studies are required to develop effective strategies for shielding more exposed struts, which are critical for achieving uniform material removal and maintaining the sample geometry integrity.

3.2.3 Post-processed geometric integrity

Figure 15 depicts the post-processed strut thicknesses for all support strategies across all inclination angles categorised into the TW and TA struts for both sample thicknesses. Although the TA struts exhibited higher strut thickness values compared with the TW struts in the as-printed condition (Fig. 12), these differences were reduced after the HPPT. Furthermore, although all average thicknesses became considerably close to the 90° reference of each support strategy for the higher inclination angles, the 45° angled TA struts did exhibit substantially less thickness reduction. The average difference of the 45° TA struts to the 90° references amounted to 73.8 μm for the 0.6 mm-thick samples and 101.5 μm for the 0.9 mm-thick samples. This finding was consistent with the lower amount of material removal, as discussed in Section 3.2.2.

Fig. 15
figure 15

Post-processed thicknesses of struts facing towards and away from the coater direction clustered by the strut inclination angle and the support strategy. a Results for the 0.6 mm-thick samples. b Results for the 0.9 mm-thick samples

Full size image

Figure 16 presents the average differences between the angled struts and the 90° references (tAngled − t 90°) under the as-printed and post-processed conditions. The HPPT diminished both the average differences and the deviation range, independent of the strut thickness, inclination angle and build orientation. However, although the support strategy did not substantially affect the as-printed strut thickness, it appeared to influence the material removal and improved the geometrical integrity of the post-processed part. Among the different support strategies, the closest alignment to the 90° reference occurred with the 0.3 mm spaced supports (S03) for the 0.6 mm-thick samples and the 0.9 mm spaced supports (S09) for the 0.9 mm-thick samples, with the maximum deviation from the 90° struts being 29.63 (0.6 mm-thick S03 samples) and 33.63 μm (0.9 mm-thick S09 samples) (Appendix 3). These support variations also showed the lowest variability between the angles, where the strut thicknesses could be controlled within the 36.19 (0.6 mm-thick S03 samples) and 37.74 μm (0.9 mm-thick S09 samples) range. For the 0.6 mm-thick samples, the 51.03 μm range of the non-supported (NS) samples was comparatively small. Considering the possibility of batch-related differences, this indicated that the HPPT can control the GDs within a 51.03 μm range, even without using additional supports.

Fig. 16
figure 16

Quantitative representation of the as-printed (AP) and post-processed (PP) average differences between the angled and 90° reference struts (tAngled − t90°). a 0.6 mm-thick samples. b 0.9 mm-thick samples

Full size image

3.2.4 Support removal

Figure 17 illustrates the post-processed samples with a 45° strut inclination for both sample thicknesses and all support strategies. Compared with the single supports at the nodes of the stent samples, support removal along the struts was insufficient, as evidenced by visible residuals in all the supported samples. The larger 0.9 mm spacing and the implementation of a four-layer gap appeared to facilitate full support removal. Although the implementation of the four-layer gap resulted in the smallest and smoothest residuals, it introduced an increased variability in the as-printed strut thicknesses, material removal and post-processed strut thicknesses (Figs. 11, 13 and 15). The support removal was also more effective in the thicker 0.9 mm samples compared with the 0.6 mm-thick samples, particularly for a smaller spacing between the supports. This suggests that a higher strut: support thickness ratio will improve the support removal because the thicker struts allow for increased current flow and higher material removal. This finding was consistent with the measured material removal data presented in Fig. 13.

Fig. 17
figure 17

Digital microscopy images of the whole post-processed 45° samples for the thicknesses of 0.6 (a–c) and 0.9 mm (d–f) and the three support strategies of S03 (a and e), S09 (b and f), S03G (c and g) and NS (d and h)

Full size image

Further investigations should focus on achieving residue-free support removal to attain full design freedom and enable the fabrication of large overhanging areas. This will require additional post-processing steps to eliminate residuals and enhance the surface quality. One possible approach for this involves extending step II while effectively shielding the sample tips (Section 3.2.1). Alternatively, other methods, such as tumbling, could be explored although they may compromise the advantages offered by the purely electrochemical hybrid post-processing approach.

4 Overall discussion and key considerations

For the first time, the present study successfully investigated the HPPT for its application on PBF-LB/M-fabricated filigree SMSs in terms of process and design. The HPPT was employed to modulate the material removal process, enabling the reduction of minimal printable strut thickness and the diminishment of the surface roughness through additional smoothing. Material removal through the HPPT was the highest for the struts exhibiting the largest GDs resulting from the low print quality and the presence of partially sintered particles. The resulting high surface roughness offered a larger area for the HPPT to act upon. Consequently, processing filigree SMSs using the HPPT resulted in an enhanced geometrical integrity of the strut thickness, independent of the strut inclination angle, build orientation or thickness. Although single supports within the part were effectively removed, the removal of the additional supports along the struts left some residues attached to the struts. The residual-free removal of these supports could have, however, enabled complete freedom of the design for the PBF-LB/M-fabricated filigree SMS containing critical overhangs. Further investigations should be performed to eliminate these residuals through additional post-processing steps. Although the HPPT has a longer processing time of up to 90 min compared with electropolishing, it offers several advantages, including the ability to combine multiple processing steps. For example, the HPPT can be used to circumvent the need for tedious manual support removal, which leads to a more efficient manufacturing process chain. Overall, The use of the HPPT exemplified by Hirtisation® offers a superior alternative to conventional methods, providing improved results in less manufacturing time and steps, addressing the intricate needs of PBF-LB/M-fabricated filigree SMS for industrial applications, particularly in producing medical implants such as stents.

Nevertheless, the results of this study highlight some key considerations that must be made to successfully apply electrochemical post-processing treatments to PBF-LB/M-manufactured parts, particularly for filigree SMSs:

  • A comprehensive understanding of the PBF-LB/M process and the effect of certain design aspects on the print quality is essential because some systematic imperfections vary depending on the machine and parameters used. Therefore, controlling critical quality factors (e.g. internal porosity) and material removal is crucial for achieving a satisfactory quality of the post-processed parts.

  • The post-processing parameters must be systematically tailored to the application to overcome batch-related systematic imperfections and challenges associated with PBF-LB/M-manufactured SMSs.

  • The presented approach demonstrated the importance of considering the whole production chain to achieve a satisfactory outcome considering that many design- and process-related dependencies are present. The overall strut allowance must be calculated with a considerable overhead.

  • The strut orientation relative to the coater direction is a critical design aspect influencing the geometric accuracy and the print quality. The strut orientation can specifically enhance the effects of compression-induced thickening and scrape-induced thinning. This phenomenon is more pronounced in the case of lower inclination angles and should be considered when applying a HPPT to filigree SMS.

5 Conclusion

In conclusion, the HPPT is a promising post-processing method for PBF-LB/M-fabricated filigree SMSs that addresses the associated challenges of dimensional process limitations, surface quality, support structures and geometric accuracy. The findings of this study contribute to enabling hybrid post-processing treatments as a viable solution for the post-processing of filigree SMSs, thereby opening new possibilities for the design and fabrication of complex and high-precision parts through PBF-LB/M. The key findings are summarised as follows:

  • ➢ Using HPPT, the strut thickness can be reduced below the given printing limit. In this study, with a given printing limit of 300 μm, a strut thickness of 150 ± 0.03 μm was achieved.

  • ➢ Using HPPT, the surface roughness can be reduced to Ra = 1.9 μm.

  • ➢ The elimination of single supports within the part through the HPPT allows for the fabrication of open-cell stent designs, which offers greater design flexibility in cardiovascular stent manufacturing.

  • ➢ The HPPT can compensate for the design- and printing-induced GDs without needing additional supports. By homogenising the strut thickness within 51.03 μm, the overall geometric integrity of PBF-LB-printed filigree SMSs can be improved.