The heating block is designed to dissipate the highest possible heat flux within the smallest possible cross-sectional area. We accommodate six 1 kW heating cartridges in a copper cylinder of 50 mm diameter. With a cross-sectional area of 1963 mm2, the heating block is therefore capable of providing a specific power density of 305 Wcm-2 at 6 kW power. The 130 mm long and 10 mm diameter heating cartridges are custom-designed by Freek GmbH (Menden, Germany) to focus heat at the tips of the cartridges. They include a 50 mm unheated area to allow the cable exit points to be distanced safely from the heat generated at the tip. The heating cartridges are placed as deep into the copper block, or as close to the tip of the melt tip, as possible (Fig. 2). The six heating cartridges form three 230 V circuits within a wye–star wired three-phase plus neutral line configuration.

Following previous melt-tip designs, we use a 60∘ cone to form the bottom of the melt-tip head to dissipate heat downwards most effectively into the ice (Talalay et al., 2019), with a parabolic shape forming the rest of the melt tip, which has worked well in previous melt tips (Kasser, 1960; Hooke, 1976). In an ideal case, the entire heat source would fit in the 60∘ cone area. We must make the copper melt tip long enough to account for the 80 mm heated length of the cartridges to avoid otherwise overheating the insulated interior of the probe. Six flat areas are removed around the widest point of the copper block to allow easy grip in a bench vice (two points) or a lathe (three points) during finishing, final tightening, and sealing.

While the design of our melt tip has already evolved over four versions during the project life, the fundamental heating block design of six 1 kW cartridges in a copper block of 50 mm diameter with a 60∘ cone remained consistent across all four versions (v0–v3; Fig. 3). The v0 was a proof of concept that simply allowed all six 1 kW cartridges to be powered up from a variable power supply but otherwise contained no electronics package. The v1 tested a polytetrafluoroethylene (PTFE) collar around the upper portion of the copper heating block to better direct heat flow downwards. The v2 discontinued the use of this PTFE collar, increased the weight considerably, and improved the sealing between the copper tip and the steel body. The final v3 increased the exposed copper heating block area outside the steel body. While our v3 form is a big improvement over our v1 form, we acknowledge that both theory and practice suggest that there is still substantial room for improvement of this copper heating block design (Shreve, 1962; Heinen et al., 2020).

Heat transfer from metal to ice is roughly 3 times as intensive as from metal to water (Kasser, 1960), so if sufficient force causes close ice–metal contact at the tip, the highest heat flux should occur within the 60∘ cone area. The remaining section of the copper, likely surrounded by meltwater, should serve to build up heat that is directed towards heat loss at the tip, as well as increasing the total amount of heat delivered into the borehole. This may result in a decrease in total efficiency, but will still likely deliver faster penetration than otherwise limiting the amount of power we can provide the tip out of fear of overheating.

The melt tip houses an electronics package that uses a custom electronics board to monitor temperatures at nine locations in the probe: four internal temperatures of the heating units, two temperatures at the top surface of the melt tip, and three locally on the circuit board. The package also tracks the acceleration and orientation recorded by a gyroscope or accelerometer (Figs. 4 and 5). We use a BNO080 triaxial accelerometer and gyroscope, which also includes a magnetometer. We install three ADS1118 ultra-small, quad-channel, 16-bit analog-to-digital converters to read the differential thermocouples of four of the six cartridge heaters plus the two independent thermocouples embedded into the copper heating block. Local temperatures, or thermocouple “cold junctions”, are measured inside the ADS1118 chips. A 12-connection screw terminal block provides a tight connection for the thermocouple leads. The thermocouple wires are shielded by metal braid, which is electrically connected to the shielding of the data cable, providing protection from electrical noise. While it would be highly desirable for the melt tip to continuously measure borehole diameter, this functionality is not yet integrated into the melt tip.

The electronics board continuously reads these six thermocouples and triaxial accelerometer or gyroscope and magnetometer values, transmitting a comma-separated serial stream over RS485 approximately every 2 s. Serial to RS485 conversion is done with an external diode for electrostatic discharge protection and 120 Ω termination resistance. Twisted pairs from the data cable are soldered directly to the top of the printed circuit board (PCB) to connect differential RS485 lines and to supply power to the PCB. For first-order noise filtering, 50 Ω resistors and 100 nF capacitors are placed on every connection to data lines, as well as a 1 M pull-up on CS pins to avoid floating during device reset. These also double as short circuit protection. Testing on early melt-tip versions indicated that all data cables, especially those linking the thermocouples with the electronics board, needed to be electrically shielded to ensure data quality by reducing electromagnetic interference associated with the cartridge heater power cables. At the topside interface, the serial stream from the melt tip is merged with the serial stream from the winch to provide a single real-time data stream for operator feedback.

The melt tip has two main sections, each housed within a separate steel body (Fig. 6). The lower section, which contains the copper heating block and electronics package, also contains an additional 2 kg copper weight block. Except for a small air cavity into which the unheated ends of the cartridge heaters protrude, the interior of the lower section is filled with a low-conductivity and high-temperature silicone. An M10 threaded rod running through the probe center from the copper block to a steel top cap is used to fasten the bottom and top components under high compression. A ∼24 mm stack of Belleville washers compressed during top cap tightening accommodates up to 5 mm of expansion associated with thermal and pressure changes.

The upper section lengthens the melt tip to permit vertical stabilization via pendulum steering under gravity (Aamot, 1970; Grzés, 1980). While pendulum steering is a traditional and reliable approach, it clearly offers no ability to steer the drill against gravity (Dachwald et al., 2014). The upper member is an unsealed and unfilled steel jacket through which the cables of the lower member pass. With five cables exiting the lower member that require waterproof connections – four cartridge heater power cables and one data cable – there is very limited free cross-sectional area around the central M10 bolt (Fig. 7). The upper section also extends the central M10 bolt to an eyebolt that serves as the structural connection to the winch cable. Both the lower and upper sections have 3 mm thick steel jackets of 50 mm diameter. Together, these members give the melt tip a total length of ∼1700 mm, with the eyebolt extending a further ∼250 mm.

The power cable, which powers the cartridge heaters, is the largest and heaviest component of the umbilical. The primary design requirement of the power cable is four conductors capable of powering the wye–star wiring of three pairs of 1 kW cartridge heaters. The secondary design requirement is the rather extreme range of operating temperatures that may be encountered. We use the ÖLFLEX Robust 210, which comprises four individually shielded fine-wire copper conductors, each with a cross-sectional area of 10 mm2. This product has a thermoplastic elastomer (TPE) outer sheath that is environmentally and chemically resistant. The TPE sheathing is rated to maintain its flexibility and integrity to a lower operating temperature of -40 ∘C.

The topside connection of the power cable is a conventional 400 V three-phase plug-and-socket connection to the interface. The bottom-side connection of the power cable is four individually crimped connections to the four power cables within the upper member of the melt tip. Each crimped connection is sealed with heat shrink tubing that contains adhesive glue. To minimize umbilical diameter, the four crimped connections are distributed over ∼1 m of power cable. This ensures that the relatively thick connections are non-overlapping and can fit within the upper member of the melt tip. While there are certainly more sophisticated and elegant methods to achieve water-resistant electrical couplings, such as pressure-rated male–female couplings, crimps and heat shrink tubes provide a relatively simple bottom-side connection that can allow melt tips to be attached under field conditions. The relevant pressure-rated connectors also have a non-trivial diameter (>25 mm) that presents a challenge to both the available cross-sectional melt-tip area and the borehole diameter.

There are no external transformers used in the power system. Conducting electricity over relatively long distances results in non-trivial power and voltage loss due to resistance. Technical specifications suggest 6.5 % voltage losses over a 500 m umbilical length at 100 % power. This results in a net resistive power loss of 386 W per 500 m and decreases the effective voltage from 230 to 215 V at the cartridge heaters. The power cable has an upper operating temperature of 80 ∘C, which means it can withstand limited resistive heating. When in air, 500 m of unspooled power cable in figure-eight formation would therefore technically require a passive environmental heat sink of at least 386 W. Resistive overheating of the power cable heating is not an issue when it is submerged in the water-filled borehole.

The data cable returns signals from the instrument package in the melt tip to the interface at the surface. The primary design requirement of the data cable was transmitting data signals in low-voltage RS485 protocol without interference from the induced electromagnetic field of the adjacent high-voltage power cable. These data signals include real-time measurements of the orientation data string and the thermocouples embedded in the cartridge heaters. For the data cable, we use an Etherline Robust PN Cat. 7 cable that has four shielded pairs of copper wire. In addition to physical shielding of the data signal, the RS485 transmission protocol makes the data signal relatively insensitive to low-frequency noise. Any 50 Hz interference from the AC power cable should influence paired RS485 signal wires in a compensating fashion.

The topside connection of the data cable is a standard male–female connector to the interface. The bottom-side connection of the data cable is an IP67 rated male–female RS485 connection. The female end of this connection is embedded in epoxy and emerges at the top of the melt tip. The male end is attached to the data cable. These multi-pin connections make the data cable connection the most complex connections of the four umbilical components. It is therefore challenging to modify data cable length under field conditions; altering data cable connections generally requires workshop conditions. The outer sheath of the data cable is TPE that is rated to maintain flexibility to a lower operating temperature of -40 ∘C. While the data cable is only used to transmit signal during drilling, it can potentially be reactivated to measure the embedded melt-tip thermocouples any time after drilling. In this way, the data cable provides a secondary mechanism for measuring ice temperature that is redundant to the thermistor cable.

The thermistor cable, which is deployed with the umbilical cord during drilling but only starts to measure and log ice temperatures once drilling has finished, represents the main sensor suite of the drill system. The primary design requirement for the thermistor cable is that, with ultralow power consumption, it can measure and log precise ice temperatures for the several months required for the thermal disturbance of drilling to slowly dissipate to background ice temperatures. The thermistor cable does not enter the melt tip. Therefore, it has no bottom-side connections; it is simply structurally attached to the umbilical with zip ties. After insertion into the ice, the topside of the thermistor cable can be connected to a solar-powered automatic data logger (Fausto et al., 2021).

The thermistor cable consists of a multi-lead wire with attached thermistors. We use an Alpha Wire multi-lead cable that has 12 individually insulated copper wires. This allows attaching 11 thermistors at desired intervals along the thermistor cable. To ensure a closer spacing of thermistors near the ice bed, where vertical gradients in ice temperature are anticipated to be greatest, the optimal spacing of these thermistors should form an exponential decay with ice depth. K-type thermistors are manually spliced into the desired positions of the multi-lead cable and then wrapped in heat shrink tubing. The outer sheath of the thermistor cable is polyvinyl chloride (PVC), which gets inflexible and brittle below -20 ∘C. While this type of thermistor cable has been used extensively by the Geological Survey of Denmark and Greenland for ice-sheet monitoring stations, digital temperature sensors would allow many more temperature measurements to be collected using a multi-lead cable with even fewer wires (Li et al., 2021a).

To recover precise ice temperatures, the thermistors are rapidly equilibrating (10 s) and the thermistor cable has no shielding that could vertically conduct heat. This lack of shielding, however, means that the thermistor cable does not return reliable data during drilling, when the adjacent power cable induces a strong electromagnetic field. The thermistors have a precision of ±1 %, or ∼0.01 ∘C, over an operating range of -80 to 150 ∘C. To maintain ultralow power consumption, the thermistor cable operates on just 2.5 V. As thermistor resistance varies as a function of ice temperature, the data signal can be very sensitive to resistive loss in the signal wire. Using a relatively large cross-sectional area (0.35 mm2) for the relatively low voltage (5 V) in the signal wire minimizes resistive voltage loss to ≪1 % over 500 m.

The structural wire bears the combined weight of the melt tip and umbilical. The primary design requirement of this wire is sufficient strength with minimum diameter. We use a steel aircraft cable that has seven bundles of 19 galvanized wires. This aircraft cable is supplied by US Cargo Control as 3/16 in. wire, equivalent to a metric diameter of 4.76 mm, and has an ultimate tensile strength of 18.6 kN. The bottom-side termination of the structural wire is a loop that passes through the eyebolt protruding from the melt tip. The loop is held by brass wire clips, which permit free movement around the eyebolt. The topside connection of the structural wire is a secure clamp on the winch drum. Once drilling has started, the length of the structural wire provides a fundamental depth limit for the drilling system.

The structural wire bears 0.448 kgm-1 of all four umbilical components (Table 1). It would clearly be desirable to have an integrated umbilical that combines these four components into a single cable (Peng et al., 2021; Zhang et al., 2021), as the combined mass of the power, data, and thermistor cables is within the structural tolerance of some off-the-shelf power cables. As previously described, however, resistive heating prevents a power cable from being spooled on a winch drum, at least without an active cooling system. As the structural wire is the only umbilical component to pass through the winch, it controls the descent rate of the melt tip and umbilical. This control function is arguably a more critical contribution to the drilling system than its structural function.

The main design requirement of the frame is bearing the load of the umbilical. We considered using an adjustable T-slot modular frame with T-nut fasteners but found that a welded steel tube structure was easier to render in the structural analysis software needed to understand how the wire load produces vertical stress in some elements of the frame and horizontal stress in other elements of the frame. The winch frame is therefore constructed with square tube steel with a thickness of 1.65 mm and welded joints. We estimate the safety factor of a welded steel frame as ∼3 when the winch is under 400 kg load. The steel frame is secured to a timber foundation of three 100mm×100mm×2000 mm cross members. This timber foundation increases stability and decreases ground pressure. Under the characteristic load of a 500 m borehole, these timbers would ultimately bear up to 315 kg (65 kg winch weight, 10 kg melt-tip weight, 15 kg foundation weight, and up to 225 kg umbilical weight) and exert a pressure of 17 kPA (2.5 psi) on the underlying snow or ice.

The length of the winch is determined by the 130 cm design requirement and the desire to provide the greatest distance possible between the encoder pulley and the winch spool. Maximizing this distance, along with using a relatively narrow but deep spool, minimizes variations in the feed angle of wire from the spool to the encoder pulley. The winch spans the borehole along its length axis, as the load is too great to cantilever over from one side of the borehole. The 800 mm width of the winch is determined by the need for a horizontal outrigger to accommodate both the spool-perpendicular motor and possible off-axis side loading when working on snow and ice surfaces. This outrigger ensures that the winch should operate with a cross-slope tolerance of up to ∼3∘. Due to space and weight considerations, the height of the winch is minimized at 560 mm.

The main design requirement of the motor is ensuring sufficiently high torque while still ensuring sufficiently small increments of pay-out from the winch spool. This requirement ultimately determined a single gear system, in which a stepper motor was paired with a 120:1 gear reduction. This ensures that each of the integer steps taken by the motor can be further reduced into a millimeter-scale unit of pay-out. As the spool radius changes through time because wire slowly pays out from the spool, the wire passes through an encoder pulley to independently measure pay-out length. During normal operation, there is no brake; the motor and gearbox together provide sufficient high-torque position control. The gearbox is the least cold-tolerant winch element, with a rated lower operating temperature of -25 ∘C. The ambient temperature of the drill tent is usually above this threshold.

The single gear system can lower or raise the wire at prescribed speeds of up to 30 mh-1, within an estimated precision of 0.01 mh-1 within the prescribed speed. The winch does not have a second higher gear for spooling the cable. At maximum speed, the motor has a maximum pulling force of 5.8 kN when all the cable is fully spooled and 11.3 kN when the cable is fully deployed. As the drilling system is intended for the one-way deployment of thermistors into the ice, the winch also does not have an auto-leveler when pulling in wire to the spool. The lack of an auto-leveler permits the winch spool to be relatively deep and narrow, reducing the set-back distance between the spool and encoder pulley. With a high gear and auto-leveler, however, the winch must be decoupled from the motor hand-spooled. We estimate that removing the winch drum from the drivetrain using a removable shaft collar and spooling a new wire takes a non-trivial 3 person hours.

The winch electronics provide the interface control over, and feedback from, the winch. They are housed in a small Pelican case attached to the frame that communicates with the interface via plain-text RS232 protocol. This communication allows the interface to prescribe winch speed (within ±30 mh-1) and log the pay-out from the encoder pulley and load from the load-pin pulley. The accuracy of the load pin is doubled by looping the wire 180∘ around the load-pin pulley. The load-pin pulley measures absolute load to within ±6 kg accuracy and relative load changes with <1 kg accuracy. The encoder pulley measures pay-out to better than ±1 mm accuracy. Under field conditions, however, the metal-on-metal combination of the encoder pulley and the structural wire allowed slippage, which substantially reduced the reliability of the digitally recorded pay-out. Manual logging ultimately provided more reliable depth estimates.

The winch motor and electronics are powered by a single 220 V cable from the common power supply. The winch motor draws a maximum of 500 W. The winch electronics draw negligible power. There are, however, separate ground lines for the motor and electronics. This protects the electronics package from any electrical noise and surges generated by the motor. In the event of an emergency, loss of power, or electrical fault, a power-cut button is depressed to engage an electromagnetic emergency brake on the motor. While there is no manual control over the winch – all instructions must be digitally transmitted via the interface – the open-source Arduino software used by the winch electronics is easy to modify and upload under fieldwork conditions. The power-cut button proved to be the easiest way to stop and restart descent during ice-sheet drilling.

The melt-tip power unit is the regulator between the raw generator power output and variable power delivered to the umbilical cable. The primary purpose of this unit is to provide operator control over power delivery within the 0 to 6 kW range, while also protecting the generator by evenly distributing load and minimizing demand spikes. The heart of the melt-tip power unit is three three-phase silicon-controlled relays (Thyro-A 3A 400-8 HRL3), which efficiently adjust the power supplied to the umbilical cable. The three 400 V line voltages from the generator are wired in a wye–star configuration, supplying 230 V across the heating elements with a common neutral line. Instead of using internal melt-tip temperature as a thermostat, regulating the generator power on/off, the relays analyze the generator's 50 Hz AC signal and deliver a percentage of each cycle as set by the operator.

The drill operator can prescribe power delivery as a percentage within the 0 to 6 kW range via a field laptop connected to the interface. The power unit logs power delivery along with all settings and modes of operation through time. These data are sent to the field laptop via USB connection using Thyro-Tool software. A key safety feature of the power unit is an independent circuit breaker (Lovato SM1R 2300) on each of the three incoming 230 V lines from the generator. These circuit breakers exist mainly to protect the operators and generator from short circuits. The silicon-controlled relays have their own circuit breakers that protect the interface in the event of a down-borehole short circuit from water ingress in either the umbilical cable or melt tip. Fuses are fitted in each relay in the event of circuit breaker failure or generator surges.

The primary task of the melt-tip data unit is receiving serial data from the umbilical data cable that transmits these data from temperature and orientation sensors embedded in the melt tip (Fig. 13). To receive these data, the melt-tip data unit must first supply the umbilical data cable with regulated 12 V to power the electronics package within the probe. This interface contains an RS485-to-USB converter, which allows the attached field laptop to monitor and log the received information using SerialStudio software. The melt-tip power unit is the regulator between the raw generator power output and variable power delivered to the umbilical cable.

The primary task of the winch control unit is to provide reconfigurable messaging to the winch, as well as receive real-time feedback from the winch. The winch has its own firmware to convert its step-motor rotation into pay-out length as a function of spool diameter. The winch control unit simply instructs this firmware of the desired speed and direction of pay-out. This instruction is transmitted via RS232 messages sent from the interface using an Arduino Mega. The winch control unit also locally displays the incoming and outgoing data stream on a small color thin-film transistor (TFT) display. This data stream includes the pay-out distance, winch speed, winch direction, load cell values, and deviation of load cell.

In the absence of a field laptop, the winch control unit of the interface has glove-friendly manual controls to select winch speed and direction. The winch control unit then converts these manually selected values into digital messages and sends them to the winch. The RS232 connection between the interface and the winch is isolated to avoid creating a ground loop. When resetting the power supply, during generator refueling for example, all interface data streams reset. This creates the largest issue for the winch data stream, as every power reset ultimately resets the winch pay-out and depth back to zero.