The submarine H.L. Hunley was the first submarine to sink an enemy ship during combat; however, the cause of its sinking has been a mystery for over 150 years. The Hunley set off a 61.2 kg (135 lb) black powder torpedo at a distance less than 5 m (16 ft) off its bow. Scaled experiments were performed that measured black powder and shock tube explosions underwater and propagation of blasts through a model ship hull. This propagation data was used in combination with archival experimental data to evaluate the risk to the crew from their own torpedo. The blast produced likely caused flexion of the ship hull to transmit the blast wave; the secondary wave transmitted inside the crew compartment was of sufficient magnitude that the calculated chances of survival were less than 16% for each crew member. The submarine drifted to its resting place after the crew died of air blast trauma within the hull.

Funding: We would like to gratefully acknowledge funding and support from the Josiah Charles Trent Memorial Foundation Endowment Fund at Duke University for directly funding this research. We would also like to acknowledge funding from the DoD SMART Scholarship Program for the research and education of Rachel M. Lance. We also gratefully acknowledge funding provided by the US Army MURI program (U Penn prime—W911NF-10-1-0526) partially supporting Cameron Bass. In addition, we would like to thank the Hagley Library’s Center for the History of Business, Technology and Society for funding via an Exploratory Research Grant that enabled the historical black powder research.

This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.

Conventional explosions can injure through one of three generic mechanisms, one of which is primary injury from the blast pressure wave itself (see Ref [ 35 ] for information on other injury types). Primary injuries from blast predominantly affect the gas-filled organs, most often causing pulmonary trauma such as hemorrhaging, but can also present as traumatic brain injuries [ 36 , 37 ]. These injury types would not be evident from skeletal or highly decomposed remains.

TNT is conventionally used as a yardstick to compare the strengths of different explosives, both low and high, which are typically described by stating their TNT relative equivalency (RE) [ 33 ]. RE is a fractional number that describes the strength of an explosive relative to the strength of TNT. Most explosives have RE values in the range 0.8–1.2 [ 33 , 34 ].

Black powder is a low explosive, a volatile blend of crushed charcoal, sulfur, and either sodium or potassium nitrate [ 28 ]. Unlike high explosives, which have burn rates faster than the speed of sound, black powder deflagrates rather than detonates. Variables such as grain size, powder density, and even the type of wood used to make the charcoal can potentially have noticeable effects on powder performance because of their impact on burn rate [ 29 – 31 ]. Tests of modern black powder have shown comparable performance in both burn rate and pressures produced to cannon-grade Union powder from the Civil War [ 29 ]. Black powder performance is also highly dependent upon the strength with which it is confined [ 30 ]. When it is spread on the ground and lit in an open, unconfined environment it burns with negligible pressure generation; however, when it is confined the charge casing allows the gradual generation of internal pressure until the point of casing failure [ 32 ]. While it is categorized as a low explosive, the data presented in this study show that it is capable of generating a sharp-rising pressure wave with certain confinement conditions.

Momentum transfer into a structure produces motion of that structure, and rapid initiation of motion can create a shock wave off the structure’s back surface [ 22 , 25 , 26 ]. Transmitted blast waves have been observed computationally behind structures and experimentally measured behind armor [ 22 , 27 ]. Computational simulations confirm that backface pressure response is the product of rapid motion of the structure wall in response to the original external blast, and because the motion response of the structure increases for stronger blasts, so will the magnitude of the transmitted blast [ 22 , 25 ]. Such a backface wave means that, even if the original blast wave is largely reflected at the front material interface of the structure, it is possible that people behind the structure could still be injured or killed by transmitted shock from a sufficiently large charge at a sufficiently close range without overt damage to the structure.

Blast interactions with structures can be prohibitively complex to test experimentally, so these tests are often performed both in air and in water by scaling down the size of the experiment according to the relevant dimensionless parameters [ 17 , 22 ]. G.I. Taylor first described the pi groups that dictate the behavior of an underwater blast wave hitting a solid structure with air behind that structure, primarily in an attempt to predict damage to ships from TNT depth charges [ 23 ]. Taylor’s dimensional analysis and subsequent studies by other groups concluded that the amount of momentum transferred into a structure by a blast wave can be scaled by size if the time-relevant parameters of the blast wave are also scaled [ 22 , 23 ]. Dimensionless parameters dictating the amount of momentum transferred are shown as Eqs ( 3 ) & ( 4 ).

The increase in both the density and the speed of sound in water means that when explosions occur underwater, the resulting shock and pressure waves travel more efficiently and further than they do in air [ 17 , 18 ]. The most critical behaviors of underwater shock waves follow traditional scaling laws since peak overpressure, duration, and impulse scale with the overall length scale of the experiment [ 17 , 19 ]. However, it is more informative and convenient to describe the output of the charge via the equations provided by Hopkinson scaling, also referred to as the principle of similitude [ 20 ]. The Hopkinson scaling equations for peak pressure and time constant for underwater blast are shown as Eqs ( 1 ) & ( 2 ) [ 17 , 21 ].

Both of two previous primary theories of sinking, suffocation and damage to the hull from arms fire, have been found to be implausible in recent publications [ 15 , 16 ]. In addition, evaluation of the attack showed that the Hunley likely drifted before finally sinking [ 16 ].

The two large holes discovered in the bow and side of the hull were determined not to be the cause of sinking by analysis of the sediment layers within, which showed that both breaches occurred long after the sinking, and no additional damage was found to the hull that provided an explanation. The holes were determined to have occurred at a later date because analysis of the types and quantity of sedimentary materials, including marine macrofauna, showed strata of sediment deposition that permitted analysis of the general patterns of sediment accumulation over time within the hull [ 13 ]. These strata indicated that the holes were not present during the vessel’s initial time underwater. The pattern of damage of the holes was determined to have been caused by a combination of galvanic corrosion, stresses from riveted seams, and erosion from ocean currents [ 14 ].

The Hunley was raised from the ocean floor in 2000, and conservation efforts have been ongoing since [ 5 , 6 ]. The skeletal remains of the crewmembers were found seated at their respective stations, with no physical injuries or apparent attempts to escape [ 7 – 9 ]. The conning towers, which formed the only path of escape, were closed with the aft tower still securely locked [ 10 ]. The bilge pumps were not set to pump out water [ 11 ]. The keel ballast weights, which could be released from within the boat, remained firmly attached [ 1 , 12 ].

The vessel’s commander could see out the fore conning tower and was responsible for navigation, while the remaining crewmen powered the vessel’s propeller from the inside using a hand crank [ 2 , 3 ]. At the other end of a hinged 16-foot spar was firmly bolted the Hunley’s torpedo, a copper torpedo of the common Singer’s design type filled with 61.2 kg (135 lbs) of black powder and fitted with a pressure-sensitive trigger ( S1 Fig )[ 4 ].

The Hunley sank the Union ship Housatonic and killed 5 Union soldiers by setting off a black powder torpedo against the ship’s hull on the evening of February 17, 1864. The narrow, tapered submarine was 12 m (40 ft) long with a maximum width of only 1.2 m (4 ft) [ 1 ]. It was shaped out of the wrought iron boiler of a previous ship, and carried a crew of 8 men ( Fig 1 ).

Methods

Scale model construction A 1/6 length-scaled model of the HL Hunley was constructed out of 16-gauge mild steel, which is materially similar to the wrought iron of the submarine’s hull in many properties including those that dictate structural response to blast exposure (S1 Table) [14, 38, 39]. Key physical design properties of the Hunley were incorporated in the scale model construction, including ballast tanks that could be filled with water and ballast weighting by lining the keel with lead. Information about the methods used to obtain the measurements of the Hunley that formed the basis of the scale model can be found in Ref [15]. The finished model is shown in Fig 2. PPT PowerPoint slide

PowerPoint slide PNG larger image

larger image TIFF original image Download: Fig 2. Photograph of the scale Hunley model, nicknamed the CSS Tiny. [a] threaded attachment for spar [b] access port (2 total, one each at bow and stern) to fill and empty the ballast tanks, can be sealed with threaded insert [c] Rings (3 on model) for carrying the vessel and attaching lines [d] Gasket-sealed panel for interior access [e] Data ports (2 on model) for gauges [f] Bulkhead fittings (4 on model) for gauge wires. https://doi.org/10.1371/journal.pone.0182244.g002 The Hunley model, nicknamed the Tiny, was exposed to underwater blasts via three primary experimental methods: shock tube exposures, black powder charges attached to the bow with a size-scaled, angled spar, and black powder charges directed at the side of the hull. For each exposure type, an omnidirectional incident pressure gauge was suspended in the center of the interior of the hull. An identical pressure gauge was also suspended in the water external to the boat, at the centerline along the length of the boat and at a distance of 6 cm horizontal standoff from the side of the hull. The gauges used were oil-filled tourmaline gauges validated for measurement of underwater and air blasts (Naval Surface Warfare Center Carderock Division, Bethesda, MD), amplified with a PCB Piezotronics 402A amplifier and powered with PCB Piezotronics model 482A10 and model 482 power supplies (PCB Piezotronics, Inc., Depew, NY). The gauge cables were foam-covered and insulated along the length of the cable between the boat and the pier. Data acquisition was performed at 1 MHz rate with 500 kHz antialiasing filters using a Hi-Techniques MeDAQ Win600e (Hi-Techniques Inc., Madison, WI). The experiment was scaled so that the degree of pressure transmission into the scaled model hull would be similar to the degree of pressure transmission into the full-sized Hunley [17, 19, 21, 40–42]. Some changes, such as the change in density from salt to fresh water, were experimentally unavoidable, but the effects are expected to be small (differences in density and bulk modulus increase the speed of sound in salt water 2–4% over that in freshwater).

Shock tube blasts The shock tube exposures were performed in the Duke University Reclamation Pond. The shock tube was composed of a driver section only, made of size 3 high-pressure stainless steel pipe flanges, fitted with a variable number of Mylar membranes (S2 Fig). The shock tube driver was braced from behind with water-saturated wooden rails and pressurized with helium until the Mylar membranes ruptured, creating a shock wave. A live charge creates spherically expanding shock waves, so because the Hunley’s spar held the torpedo at a downward angle the Hunley would have been directly exposed from all sides along the entire length of its hull. Since shock tubes create highly directional shock waves rather than spherical shock waves like those produced by live charges, the shock tube was used to characterize which sections of the hull were responsible for transmitting the effects of blast. The characterization was performed by directing the shock tube at the bow of the vessel, at the side of the vessel, at the vessel’s keel, and at oblique angles relative to the axis of the hull. Once it was determined that the bow of the vessel transmitted negligible blast effects into the main cabin, and the perpendicular component of the blast was responsible for blast transmission into the cabin, the transmission tests were performed by directing the shock tube perpendicularly at the side of the hull. The external pressures were estimated to be the perpendicular component of a blast with the correct direction of propagation, and so were divided by sin(11°) to calculate the overall peak pressure values of the estimated blast [43].

Live charge blasts The test site for the live charges was a freshwater pond with a bottom depth slightly greater than the scaled value (9 m/6 = 1.3 m) of the bottom depth at the location of the Hunley attack on the Housatonic [44]. This depth would ensure that reflections of the blast waveform off the bottom would be equal to or less than those experienced by the Hunley, and so would either approximate the amount of bottom reflection or err on the low side since bottom reflections augment the strength of an underwater blast exposure [17]. All necessary permits and legal permissions were obtained prior to each round of testing. Charges were packed with 4Fg black powder (Goex Powder, Inc., Minden, LA) with casings constructed out of schedule 80 PVC pipe with threaded end caps. The historical drawing of the Hunley torpedo indicates that it was filled with grade FF cannon powder (S1 Fig). However, samples of powder from recently uncovered Union cannonballs from the Civil War indicate that the historic FF grain size more accurately matches the modern 4F grain size standard for musket powder [29]. The charges were triggered using NPb squibs (Martinez Specialties, Groton, NY). Isolated squibs were set off to evaluate their detonation signatures at the distances of interest, but were determined not to have a noticeable impact on the pressure waveforms. Several preliminary underwater tests were conducted with variations in charge size, casing construction, and range to the point of measurement to assess black powder’s performance with respect to the scaling of time constant for the non-dimensional groups. Charge size was varied, with sizes of 283 g, 455 g, 490 g, and 1 kg, and range between the charge and the point of measurement was varied between 80 cm and 1.8 m. The range values were selected to validate the scaling principles in the regions most relevant to the Hunley. Initial testing showed that orientation and position of the charge relative to the gauges had a measurable effect on the pressure waveforms. Therefore, test data were only used for the analysis of scaling if they had the same charge orientation and depth as the tests with the scale boat model, and gauge locations that would fall along the length of the submarine hull. Specifically, the tests evaluating the effects of confinement strength were eliminated from the scaling data set because the measurements were taken at the same depth as the charges. The initial time constant of decay (θ) was measured for all blasts. Theta was scaled using Hopkinson scaling by division by the cube root of charge weight (W1/3) (Eq (2)). This value was then plotted as a function of the scaled distance (W1/3/Range) at which the waveform was measured. A power law equation was fitted to the data in the scaling data set using least-squares regression. This method is the standard procedure for describing the time-scaling behavior of explosives and is referred to as the principle of shock wave similitude [17, 21]. The Tiny model was blasted with black powder charges of three sizes: 283 g and 455 g charges, corresponding to 1/6 and 1/5 size scale of the 61.4 kg (135 lb) Hunley torpedo, and 1 kg charges, which were the maximum size as requested by the ATF. While 283 g is the properly mass-scaled value for black powder, the larger charge sizes were constructed to evaluate the degree of propagation of higher pressures through the hull wall. Experimental limitations on scaling burn rate of the powder and methods of charge confinement meant that the PVC 283 g charges would severely underestimate the strength of the exposure compared to the original copper-cased torpedo; larger charges were therefore also tested to evaluate how the transmission properties changed with increases in external pressure. All charges except one were attached via a size-scaled spar to position them in the same manner as the Hunley’s torpedo. One 283 g charge was positioned beside the boat to further increase external pressure of exposure. The external pressure from this charge was divided by the sine of the angle between the direction of blast propagation and horizontal at the centerline of the keel (11°) to calculate the total external peak pressure from a spar-mounted charge that would have the same amount of propagation through the hull. This angular correction for direction of transmission is often used in structural shock testing for charges in different geometric orientations from their targets [43].

Shock tube blasting of metal plate Propagation of a sharp-rising shock wave through the full-sized Hunley structure was investigated using a mild steel plate with greater thickness (1.6 cm, 5/8”) than the original Hunley hull (1.0 cm, 3/8”). The purpose of this test was to ensure that the shock wave maintained the sharp rise time critical to cause injuries even when propagating through a material at least the thickness of the full-sized submarine hull. The steel plate was a square 61 cm (24”) on each side and was exposed to airblast using a helium-driven shock tube. The shock tube was 30.5 cm (12”) in diameter and was aimed at the center of the plate. A standoff distance of 4 cm was set between the plate and the end of the tube to allow lateral venting of the shock and provide reduced impulse on the plate relative to peak incident pressure. Incident pressure was measured at the end of the shock tube using 200 psia Endevco pressure gauges (Model 8530B-200, Meggitt Sensing Systems, Irvine, CA) that were flush with the internal wall of the tube body. Two additional Endevco pressure gauges were rigidly fixed behind the center of the steel plate, with 10 cm between the back of the plate and the center of the gauge faces. Both gauges were oriented to measure incident pressure.