MeV x-ray production from a petawatt laser in the regime of a relativistically transparent preplasma, with applications to radiography Open Access

Mega-electron volt (MeV) x-ray sources are essential tools for nuclear physics,¹ medicine,² flash radiography,³ and computed tomography of high-Z objects.⁴ As a consequence of their straightforward principles of operation, bremsstrahlung sources are often chosen for many applications. A conventional bremsstrahlung source consists of a source of MeV electrons, generated via an accelerator or pulsed power machine, after which this beam traverses a mm-scale metal converter. After sweeping electrons away with a deflection magnet, the MeV x-rays proceed in the forward direction, available in isolation for the desired application. However, conventional sources remain limited in their large source size (⁠ $∼$mm) and pulse duration (⁠ $∼$ 100 ns).^3,5

When a Petawatt (PW)-class laser is incident on a similar converter, such as a mm-thick metal target, the laser can accelerate electrons from the interaction interface along the laser direction.⁶ When these electrons, reaching tens of MeV, stream through the target, they radiate x-rays in a broadband continuum up to the maximum electron energy,⁷ with percent-level conversion efficiency.⁸ Due to the ultrafast nature (⁠ $∼$ps) and tight focus (⁠ $∼μ$m) of high peak power lasers, the pulse duration and source size are quite competitive with conventional sources. Laser-driven sources have pulse durations no greater than tens of picosecond⁹ in the MeV range, with source sizes down to tens of micrometer.¹⁰ In addition, the half-angle divergence of the MeV x-rays, typically 10°–20° for a laser–solid interaction, is controllable as it depends on the f/# of the optic.^8,10

Despite these benefits, no laser system has been constructed for the explicit purpose of delivering an MeV x-ray source, in part due to the risks of unknown target physics. Though the x-ray dose output for high flux applications, such as radiography, can be met with a laser,¹¹ few studies have avoided the instabilities that compromise stable dose for a statistically significant number of shots. Though previous studies have shed light into the role of a preplasma on MeV x-ray production,^12–16 shot-to-shot variation often exceeds an order of magnitude^17,18 from mechanisms not explicitly attributed to, but likely, instabilities such as laser hosing.¹⁹ Though this may be avoided by using high-contrast lasers,²⁰ suppressing shot-to-shot variation would allow readily available lasers of modest contrast to deliver sources of immediate utility.

Multiple experimental techniques are expected to reduce the shot-to-shot variation in MeV electron and x-ray spectrum and yield. First, in situ confocal microscopy can enable target positioning within micrometers of best laser focus,²¹ thus reducing shot-to-shot variation from target alignment. Second, reaching high laser intensity with a small f/# parabola leads to a larger electron and thus x-ray divergence,^8,10 which spreads the radiation over a larger solid angle, thus smoothing out any variations in the beam profile from shot-to-shot. Third, if the laser contrast is sufficiently modest, the scale length of the high-density (⁠ $ne/nc>1$⁠) preplasma will be large enough to improve laser coupling to MeV electrons, while being small enough to prevent growth from the aforementioned laser hosing instabilities.¹⁹ 

Here, we use the Texas Petawatt laser²² to implement the three techniques mentioned above, generating a highly reproducible source of MeV x-rays. The laser, focused with an f/1.5 off-axis parabola (OAP), reaches a peak intensity on target of $1021$ W cm⁻² (⁠ $a0≈27)$⁠, where the laser focal plane is aligned to the target with $∼μ$m precision. The modest contrast laser, measured at the time of our campaign to be 10⁷ at 50 ps before the main pulse (see the supplementary material), generates a preplasma. This preplasma contains both low-density (⁠ $ne/nc<1$⁠) and high-density (⁠ $ne/nc>1$⁠) components before the arrival of the laser peak. In particular, the overcritical plasma, in which the MeV electrons are largely accelerated, is numerically predicted to be only $∼$ 3  $μ$m in length. This length of overcritical plasma is too short for hosing instability growth,¹⁹ but sufficiently large for the laser to accelerate electrons in this volume. By changing the target material in the interaction region from a high-Z metal (W or Ta) to plastic, an MeV x-ray dose increase of 60% was observed, with particular enhancement in the high-energy tail of the x-ray spectrum (> 5 MeV). In this tail, coating the target with plastic improves the dose by more than a factor of 2. Hydrodynamic simulations via xRage²³ predict similar preplasma scale lengths between the bare metal and plastic-coated targets, for both the low-density and high-density preplasma components. By initializing kinetic simulations with these preplasma profiles, VPIC simulations reveal that the plastic (idealized as the hydrocarbon CH) preplasma in the initially over-dense region can more readily undergo relativistic transparency.^10,24–26 Because of the greater sound speed and smaller ion inertia of the plastic, lighter plasma ions can move more rapidly in response to the laser $J×B$ and ponderomotive forces. This faster response time ultimately leads to a larger volume of dense plasma interacting volumetrically with the laser pulse, enabling self-focusing and improving laser energy coupling to electrons. Because this enhancement relies on changing the preplasma material and not the density scale length, this enhancement is highly reproducible, as long scale lengths are associated with instability growth.^12,16 The demonstration of < 10% shot-to-shot variation in x-ray dose is beneficial for radiography and other applications.

The experiment was carried out at the Texas Petawatt (TPW) Laser Facility, in which the 1.057  $μ$m laser is focused down to a $∼$ 3  $μ$m radius focal spot (containing 50% of the laser energy) via an f/1.5 off-axis parabola. With this final focusing optic, the laser was able to deliver peak intensities of $∼1021$ W cm⁻², with 120 J in a 140 fs FWHM pulse, which is linearly polarized in the vertical direction. To avoid strong back reflections returning up the laser chain, laser incidence on-target is 45°, tilted horizontally. With a confocal microscope, the front surface of the target was precisely aligned within $±$ 2  $μ$m of best laser focus.

Figure 1 illustrates the two configurations of the experiment, with key experimental parameters summarized in Table I. Configuration I sought to determine a possible material dependence on MeV electron generation, as these are the electrons of interest for generating MeV x-rays. In order to avoid full transparency or excessive MeV electron loss in the target, micron-scale targets were chosen. To measure these electrons, the Global Spectrometer for Positron and Electron Characterization (GSPEC)²⁷ was fielded along laser axis. As detailed in Sec. III, configuration I allowed us to identify superior production of MeV electrons from 1  $μ$m of plastic (mylar) vs 5  $μ$m metal (W).

Once a target material dependence was demonstrated for electrons, configuration II sought to measure how these differences in electron sources for different laser target materials would manifest as differences in the MeV x-ray sources produced. To convert MeV electrons into x-rays, mm-thick metal (Ta) targets were either bare or coated with $2 μ$m of plastic (Parylene-N). The MeV x-rays were measured with a filter stack spectrometer (FSS) along the laser axis, downstream from a deflection magnet to minimize background noise from charged particles. The FSS consists of a W housing with a 5 mm collimator, with 20 layers of 25 × 25 mm BAS-SR image plates behind filters of increasing Z, though the $∼$ 500  $μ$m image plates contain enough material²⁸ where their filtering behavior must also be accounted for. Table II shows the stack configuration excluding the image plates between each layer. The low-Z filters have a monotonic decrease in attenuation for multi-MeV photons, while the higher-Z filters produce significant Compton electrons that assist with spectral reconstruction. Both the image plates used in the FSS and GSPEC were scanned and converted into units of photostimulated luminescence, or PSL, via a General Electric Typhoon FLA 7000 IP scanner. Once in PSL, image plates calibrations were applied to get absolute particle counts for electrons²⁹ and photons.^10,30 In addition to measuring the x-ray spectrum, an x-ray radiograph demonstration platform was also fielded on configuration II. The platform consists of the kaleidoscope resolution target in front of a Varex 4343 flat panel detector to generate a digital radiograph. The kaleidoscope resolution test object,^10,15 17 mm thick and 80 mm in diameter, has feature sizes ranging from 0.1–4 mm for measuring the resolving power of a radiography system.

From the laser interaction with micrometer foils (configuration I), a substantial difference was observed between 5  $μ$m W and 1  $μ$m CH foils, especially in the high-energy component [Fig. 2(a)]. When fit to a simple exponential, a decay constant, or “temperature” of the hot electrons, can be extracted. Both targets produced temperatures of $∼$ 3 MeV for the lower energy component distribution (⁠ $≲$7 MeV), though the yield of electrons in this component increases by $∼$70% with plastic foils. However, for the high-energy tail, electron temperature and yield are both $2×$ higher from the CH target. The relative improvement in MeV electron injection into the target is likely underestimated with this measurement, as a result of the strong electrostatic sheath field that forms on the rear surface. This sheath field is stronger for thinner foils, such as the 1  $μ$m CH foil studied here, and it modifies the electron spectra because not all electrons driven into the target can escape.^31–33 

This difference in target type motivated our follow-on study using configuration II, in which we compared MeV x-ray production from bare 1 mm Ta targets vs 1 mm Ta coated with $∼$ 2  $μ$m of plastic. Due to materials availability, the micrometer foils and thick targets are similar but not an exact match in composition. The plastic foil is mylar, while the plastic coating is Parylene-N, and the metal foil is tungsten (Z = 74), while the mm target is tantalum (Z = 73). However, the materials are similar enough to use the micrometer foil experiment to guide the mm target experiment.

In switching to configuration II, the filter stack spectrometer (FSS) replaced the GSPEC to measure laser-axis radiation. The FSS was configured to be sensitive to photons > 30 keV, with the photon beam depositing measurable amounts of energy on all 20 layers. Figure 2(b) shows the energy deposited by x-rays on each layer, with a clear difference in energy deposition from bare Ta targets vs targets coated with plastic. The PSL values do not decrease monotonically; rather, the “jumps” in PSL correspond well to the material transitions of increasing Z. In the stack, layers 6–7 transition from plastic to aluminum filters, and layers 14–15, from aluminum to lead (Table II). As a result of the sharp gradient in electron number density within the material, substantially more Compton electrons are generated, and MeV electrons deposit their energy $∼$ 100  $×$ more efficiently in image plates when compared to MeV photons.^10,30 Thus, Compton scattering and other secondary production processes must be accounted for in calculating the FSS response matrix.

The raw data of Fig. 2(b) were used to generate the spectral inversions of Fig. 2(c) via the unfolding method described in Wong et al.³⁴ The inversion requires a response matrix generated for the spectrometer in a radiation transport code, in this case Monte Carlo N-Particle (MCNP)³⁵ (see the supplementary material). This method is based on perturbative minimization and requires no assumption of spectral shape or features, only that the spectrum is generally smooth. As detailed in Wong et al., the unfold progresses by analyzing spectra at increasingly high energy resolution to minimize artificial noise. The candidate spectra are smoothed every 10 iterations until a smooth spectrum is recovered with 30 energy bins. The spectra shown in Fig. 2(c) are averaged over 3 unfolds, with few-percent deviation for each unfold, where the total x-ray energy is generally conserved. Because of experimental noise levels around 10 mPSL, we are limited in our ability to measure counts below 10⁸ photons/MeV/sr; hence, we cut off our spectrum at this point. From the electron spectra [Fig. 2(a)] is likely that x-rays > 25 MeV are also produced, but in yields too low for the FSS to detect.

Similar to the electron spectra, the first x-ray temperature component remains similar (⁠ $∼$ 800 keV) regardless of whether plastic is present, though dose is increased in this component by $∼$ 50%. For MeV x-ray sources, the output is often characterized by the dose, typically reported in units of Rad@1m. 1 Rad@1m is equivalent to depositing 10 mJ/kg of x-ray energy in a slab of dry air 1 m from the source.¹⁰ For the high energy component of the spectrum (>5 MeV), the dose increases by 2.2  $×$ with the plastic coating. The boost in dose from the entire spectrum is 60%, which is replicated in MCNP when simulating the injection of the Fig. 2(a) electron spectra into a 1 mm Ta converter. However, there are not two distinct low and high energy spectral components in MCNP (see the supplementary material), likely due to MCNP not containing the physics of target heating and plasma sheath formation. Though constraining the limits of Monte Carlo transport codes is beyond the scope of this paper, the agreement in the relative dose output with MCNP supports a previous work that escaping electron spectra may approximate the spectral shape to the injected electron spectra.³¹ However, it is important to note that this does not account for the presence of the electrostatic sheath field, which may modify the spectral shape, and decrease the number of MeV electrons by an order of magnitude.³² In addition, as the sheath field is stronger for thinner targets,³³ greater modification in spectra is expected for the 1  $μ$m CH foil relative to the 5  $μ$m W foil. Despite not being able to capture all of the physics of radiation transport within the evolving target, the trend in enhancing MeV radiation generation in CH by $∼$ 60% is evident through multiple studies: simple foils, mm-thick targets, and MCNP modeling with the measured electrons. Table III summarizes our observations on the increase in radiation production enabled by plastic in the laser–solid interaction region.

In VPIC^36–39 simulations, we model the laser pulse as a two-dimensional (2D) normally incident s-polarized Gaussian beam with a $sin2$ temporal pulse focused on the initial target surface. As noted by Stark et al.,⁴⁰ cross-comparisons with 3D simulations show that 2D s-polarized laser conditions provide the most physically accurate, computationally tractable means of modeling the laser–plasma interaction in 2D. To approximate the on-shot laser conditions, we use a Gaussian laser beam with the f/# and intensity modified to provide a similar laser energy while accounting for oblique incidence and expected target positioning jitter. Our simulations have been performed with characteristic pre-plasma conditions obtained from the rad-hydro code xRage (see  Appendix for details). We use two exponential functions to approximate the pre-plasma: one from peak density (253  $ncr$⁠) to $ncr$ and a second, for density below $ncr$⁠. The characteristic scale-lengths are $Ln1=0.5$ and $Ln2=7.4 μ$m. We caution that owing to experimental and computational limitations, one cannot realistically expect quantitative agreement between the kinetic modeling and the measurements. Instead, we seek from our modeling to gain an understanding of the governing physics in order to understand better the present experiments and to guide future experiments.

We report on simulation results below using an f/5 focal geometry and an intensity of $9×1020$ W cm⁻², contrasting the behavior of CH coated W vs bare W targets. Figure 3(a) shows the time-integrated electron spectra measured on the simulation boundaries for the CH+W and bare W targets. The time $t=0$ is defined at the initiation of laser injection from the left boundary. Owing to the crossing time for the laser accelerated electrons traveling from the target to the boundaries, the electron spectrum shown at a time represents electrons accelerated at earlier times in the simulations (note that relativistic electrons require $∼$ 100 fs to travel 30  $μ$m). The spectra are obtained at the ends of the simulations at $t=690$ fs, which is 5  $×$ longer than the FWHM laser pulse length in order to capture the laser-accelerated electrons arriving at the right boundary. Overall, the forward MeV electron production increases by $∼$ 60% by coating the W target with plastic, in good agreement with the experiment. By collecting the MeV electrons reaching the right boundary at the end of the simulation, we estimate a $∼$ 13% conversion efficiency from the laser to electrons for the W target, compared to $∼$ 20% with plastic. Though the numerical simulations do not recover the spectral details of the experiment, the $∼$ 60% improvement enabled by plastic supports the experimental observations (Fig. 2).

Insight into the target dynamics is gained via one-dimensional (1D) electron density lineouts for CH+W and W targets during the laser peak at $t=200$ fs [Fig. 3(b)], and at the trailing end of the pulse at 380 fs [Fig. 3(c)]. The lineouts, centered on the laser axis, are averaged over a 0.16  $μ$m width. In both frames, the dashed curve shows $ne/ncr$ at $t=130$ fs when the pulse just begins to interact with the over-dense layer, representative of the initial density profile prior to the main pulse, the same for both targets. As time progresses, $J×B$ and ponderomotive forces from the laser compress the target electrons in the over-dense region, leading to sharp density pile-ups in both targets. This creates a charge imbalance and an electrostatic field, which pulls the ions forward to follow the electron motion. However, the higher sound speed and the smaller ion inertia in the CH+W target lead to a larger density pileup, which also travels ahead of the density pileup in the W target. For the CH+W target, the laser can penetrate deeper into the over-dense plasma from relativistic transparency effects^{25,26,41–43} and the turning point of the laser (i.e., at the sharp density rise of the pileup) is ahead of that for the W target by more than a micrometer at $t=380$ fs, when the laser pulse is largely over. The net effect is that in the CH+W target the laser is able to volumetrically accelerate a larger number of electrons via relativistic transparency to higher energies from a combination of two effects. First, electrons in a larger volume of the target plasma are displaced by the laser; and second, there are more electrons residing in the final skin-depth layer near the (relativistically modified) critical density for CH+W targets compared with W targets. As xRage may be limited in its ability to capture the near-relativistic electron motion driven by the TPW pedestal (see the supplementary material), we caution that calculations of the low-density $Ln2=7.4$ and high-density $Ln1=0.5 μ$m scale lengths may not perfectly match the experiment. However, VPIC simulations, which were initialized using these scale lengths, indicate that the difference in target material in the same scale length preplasma is sufficient to explain the $∼$ 60% enhancement observed experimentally.

As the electrons in the high-density preplasma become displaced by the rising laser pulse (200 fs), the laser is observed to self-focus in the high-density preplasma [Fig. 3(b)]. The effect is greater for the CH+W target, where the laser field doubles in the high-density region. This effect is more modest in the high-density W preplasma, where the laser field increases by 65%. These correspond to increases in the peak laser intensity by $2.7×$ for the W foil, compared to $4×$ for the CH+W foil [Fig. 3(d)]. To achieve the higher laser field (and thus higher intensity), the laser spot self-focuses to an $∼$ 80% smaller spot size at the pileup region for the CH+W target relative to the W target. This modest difference in spot size (1.5  $μ$m in the CH+W pileup region vs 1.8  $μ$m in the W pileup region) does not significantly change the focusing geometry, as evident in the 2D laser field profile in the simulation box [Figs. 3(e) and 3(f)]. Because the focusing geometry in the high-density preplasma is similar for both target types, we do not expect a significant difference in radiated electron and x-ray profile.

High quality x-ray transmission radiographs of high areal density objects (tens of grams square centimeter or greater) require MeV x-ray sources to penetrate the object sufficiently to form an image. These sources must also be highly reproducible,⁴ as discrepancies between spectra and yield introduce challenges in generating images systematically, important for probing single-shot dynamic experiments and for radiography on a production line. To the knowledge of the authors, the spectra of most conventional MeV x-ray sources are not measured routinely, at least across the U.S. Department of Energy complex. Rather than measuring the MeV x-ray spectrum on most shots, which requires sophisticated analysis, most conventional sources rely on radiation calorimeters⁴⁴ to evaluate source stability. With their instant readout, radiation calorimeters provide single-shot dose for pulsed sources, or dose rate for continuous sources. Though the doses we report here (2d) are well below the > 10 R@1m produced by a 2.4 kJ short pulse laser,¹¹ we have demonstrated a high degree of source stability in the MeV x-ray range, as required for many applications. Across 17 bare 1 mm Ta shots and 4 plastic-coated Ta shots, we show that dose measurements from Texas Petawatt vary < 10% per shot, which is sufficiently reproducible for a radiography facility.

Another important consideration for a radiography facility is the resolving power of the system, often constrained by source size. For conventional MeV x-ray sources, the source size is on the order of a millimeter,¹⁰ which can greatly limit the ability to recover fine features in an x-ray transmission radiograph. To achieve sub-millimeter magnification with a millimeter focal spot, a multi-meter standoff distance is required, and building a $∼$ 10 m shielded radiography bay costs >$10M. Shielding costs are lower with a microfocus geometry, and high power lasers have demonstrated source sizes of < 100  $μ$m with a bremsstrahlung interaction.¹⁰ Few-hundred-microns, as measured here at the Texas Petawatt (Fig. 4), is typical and is still quite competitive with the $∼$ mm sizes from conventional sources (see Wang et al.⁹ and references therein).

Bright sources of MeV x-rays are a cornerstone of nuclear physics, oncology, and radiographic imaging. Here, the generation of MeV x-rays from a PW laser was investigated experimentally, with an ability to enhance x-ray generation > 0.1 MeV by $∼$ 60% by coating mm-thick Ta targets with 2  $μ$m of plastic. The mechanism for enhancement relies on a generating a preplasma with both low-density (⁠ $ne/nc<1$⁠) and high-density (⁠ $ne/nc>1$⁠) components. VPIC simulations indicate that due to the greater sound speed and lower ion inertia in the plastic, the plastic more readily undergoes relativistic transparency, improving the production of electrons up to tens of MeV. These additional electrons traverse mm-thick Ta to produce the MeV x-rays via the bremsstrahlung process, with the entire system demonstrating a resolving power of < 300  $μ$m. This system is also highly reproducible, with < 10% shot-to-shot variation in x-ray dose for both bare metal and plastic-coated targets. Though the detailed evaluation of the stability mechanism is beyond the scope of this work, we hypothesize that there are three major contributors. First, the precise target alignment with a confocal microscope enables target placement within $±$ 2  $μ$m of best laser focus. Second, the small f/# optic (f/1.5 in this case) leads to a large divergence of MeV electrons and x-rays.^8,10 In this system, any electron or x-ray pointing fluctuations of a few degrees will not be pronounced on the GSPEC or FSS detectors, as the large divergence smooths out any variations in electron and x-ray beam profile from shot-to-shot. Third, with a modest laser contrast (10⁷ at 50 ps in this case), a steep overcritical preplasma is generated on the front surface. Our xRage modeling predicts that this high-density preplasma component is only $∼$ 3  $μ$m long. This volume is deep enough for the laser to volumetrically interact with the preplasma without giving rise to instabilities such as laser hosing.¹⁹ 

The source is reproducible for both plastic and metal in the interaction interface. We believe that changing the material to improve coupling, rather than increasing the scale length,^12,16 may suppress instability growth, though this is a topic of future numerical investigation. For this work, one VPIC simulation was run for each target case due to computational limitations. However, additional VPIC simulations, with random seeds in the preplasma, are of interest as they may provide insight into the narrow shot-to-shot variation.

For commissioning a facility for laser-driven radiography, predictive simulation codes should inform the user of the spectral output and dose for arbitrary laser and target parameters. However, significant gaps in predictive capability remain. One major contributor to this is the difficulty in measuring the preplasma generated before the arrival of the main pulse.^12,45 Not being able to probe the preplasma on most experiments requires heavy reliance on hydrodynamic codes to simulate preplasma expansion. Thus, benchmarking xRage predictions by probing the preplasma is a topic of future work. Though similar studies have been conducted,^12,16 we aim to broaden this dataset using repetition-rated targetry, inspired by those in the literature.^46,47 In addition to the plastic and metal targets studied here, targets would include critical-density foams,⁴⁸ which have been shown experimentally to enhance MeV electron and x-ray production through relativistic transparency.⁴⁸ In our work, we observe the coating delaminate on mm length scales, which would, for a repetition-rated Ta target wheel, increase target replacement by an estimated $∼$ 10  $×$⁠. However, we continue to pursue coated targets because maximizing dose at expense of increased target replacement is important for single-shot high-dose applications, such as flash radiography. For all target types, we aim to couple preplasma measurements with angularly resolved electron and x-ray spectra. For this purpose, scintillator-based x-ray spectrometers, similar to those described in the literature,^49,50 are under development. Constraining the simulations with angularly resolved energy partition, as well as target expansion, will improve the predictive power of simulations, thus providing guidance on a next-generation laser facility built for radiography.

See the supplementary material for the following: the response function for each image plate layer of the filter stack spectrometer (FSS) (Figure S1); simulated MeV x-ray spectra generated in MCNP with a 1 mm Ta converter, using the measured electrons from Fig. 2(a) as the input (Figure S2); and the laser intensity contrast measured at the time of the experiment (Figure S3).

This work was performed under the auspices of the U.S. Department of Energy by Triad National Security, LLC, operator of the Los Alamos National Laboratory (LANL) under Contract No. 89233218CNA000001, with support from the Laboratory Directed Research and Development (LDRD) Program. Computing resources were provided by the Advanced Technology Computing Campaign (ATCC) and LANL Institutional Computing programs. Beamtime at the Texas Petawatt laser was provided by LaserNetUS via the DOE, Office of Science, Fusion Energy Sciences, under Contract No. DE-SC0021125.

The authors have no conflicts to disclose.

J. Strehlow: Conceptualization (supporting); Data curation (lead); Formal analysis (lead); Investigation (equal); Methodology (supporting); Supervision (supporting); Writing – original draft (lead); Writing – review & editing (lead). L. Yin: Formal analysis (equal); Investigation (equal); Writing – original draft (equal); Writing – review & editing (equal). C.-S. Wong: Formal analysis (equal); Investigation (equal); Methodology (equal); Software (lead); Writing – review & editing (supporting). S. V. Luedtke: Methodology (equal); Software (equal); Writing – review & editing (supporting). S. Palaniyappan: Conceptualization (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Project administration (supporting); Supervision (equal); Writing – review & editing (supporting). D. J. Stark: Formal analysis (equal); Investigation (equal); Validation (equal); Writing – review & editing (supporting). C. K. Huang: Investigation (supporting); Methodology (equal); Writing – review & editing (supporting). A. Bogale: Data curation (equal); Investigation (equal). B. Cage: Data curation (equal); Investigation (supporting). T. A. Coffman: Data curation (supporting); Investigation (supporting). A. Figueroa Bengoa: Formal analysis (equal); Investigation (equal). R. Fitzgarrald: Data curation (equal); Investigation (equal). L. T. Mix: Data curation (equal); Investigation (equal); Methodology (supporting). R. Nedbailo: Data curation (equal); Investigation (equal); Methodology (supporting). D. R. Rusby: Data curation (equal); Investigation (equal); Methodology (supporting); Writing – review & editing (equal). J. L. Schmidt: Data curation (equal); Investigation (supporting). J. Twardowski: Data curation (equal); Formal analysis (supporting); Investigation (equal). A. Van Pelt: Data curation (equal); Investigation (equal); Methodology (supporting). T. H. Day: Investigation (supporting); Resources (equal). B. J. Jones: Investigation (supporting); Resources (equal). S. A. Bruce: Data curation (supporting); Resources (equal). A. Helal: Data curation (supporting); Resources (supporting). M. M. Spinks: Data curation (equal); Resources (supporting); Writing – review & editing (supporting). H. J. Quevedo: Data curation (supporting); Investigation (equal); Methodology (supporting); Writing – review & editing (equal). F. N. Beg: Funding acquisition (supporting); Project administration (supporting); Supervision (equal). E. A. Chowdhury: Supervision (equal). T. Ditmire: Project administration (supporting); Resources (supporting); Supervision (supporting). E. Liang: Investigation (supporting); Methodology (supporting); Supervision (equal); Writing – review & editing (supporting). A. G.R. Thomas: Supervision (equal); Writing – review & editing (equal). J. C. Fernandez: Conceptualization (equal); Investigation (supporting); Methodology (equal); Supervision (supporting). D. C. Gautier: Conceptualization (equal); Data curation (equal); Investigation (equal); Methodology (equal); Supervision (equal). J. Hunter: Conceptualization (equal); Investigation (supporting); Methodology (equal); Supervision (supporting). Y, Kim: Data curation (supporting); Investigation (supporting); Methodology (supporting); Supervision (supporting). K. D. Meaney: Data curation (supporting); Investigation (supporting); Methodology (supporting); Supervision (supporting). B. J. Albright: Conceptualization (equal); Funding acquisition (lead); Investigation (lead); Methodology (lead); Project administration (lead); Resources (lead); Supervision (lead); Writing – review & editing (equal).

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

Radiation hydrodynamic simulations of the target expansion due to the prepulse were performed using 1D xRage. Over the span of 100 ps prior to arrival of the main pulse, we simulated the laser incident onto a W target slab of 20 μm in depth, with and without additional CH plastic layers in front. xRage uses adaptive mesh refinement, importing tabular equation of state data from the SESAME⁵¹ library. Material opacity data are generated from TOPS.⁵² A convergence study on cell size was performed to ensure appropriate resolution, and a 10 nm resolution was used for all simulations. The default inverse bremsstrahlung absorption mechanism was used,⁵³ but additional testing for fixed absorption values produced qualitatively similar expansion features. Preplasma scale lengths calculated in xRage were then imported into VPIC to study the interaction of the main pulse with the plasma.

The VPIC simulation dimensions are 41 and 24  $μ$m in the $(x,z)$ plane with the laser launched from the left boundary propagating in the x direction. The CH and W layers have the same peak density 253  $ncr$ and are 1 and 2  $μ$m thick for the CH+W target, and 1  $μ$m thick for the W target. The location of the high-density target surface adjacent to the exponential pre-plasma is at $x=36 μ$m where the laser is focused on. We use 20 500 and 12 000 cells in the x and z directions (with 0.98 Courant condition), which correspond to over 5 cells per skin depth of the peak density region to allow for more accurate description of the laser–plasma interaction dynamics during relativistic transparency. There were either 64 or 512 particles per cell per species, as testing on some simulations produces effectively identical results. The initial temperature of the electrons is 15 keV, chosen to ensure that the plasma Debye length is adequately represented in the simulation for numerical convergence.⁵⁴