The dabba in your bag, a thermos and a spacesuit are the same machine at three scales: each stacks barriers against the three ways heat moves, which are conduction, convection and radiation. A plain steel dabba blocks almost nothing and, by my estimate, with these example inputs (400 g of food, 75°C, a 30°C room), drops from 75°C to below the 60°C hot-food line in about 35 minutes. A vacuum lunch jar removes conduction and convection and takes nearly 4 hours, and a NASA suit adds layers of shiny film in a vacuum to cut radiation by a few hundred times.
- Three paths: heat leaves by conduction, convection and radiation; every insulator blocks some of them.
- About 35 minutes: my estimate, with these example inputs, for a plain steel dabba to fall from 75°C to 60°C in a 30°C room.
- Above 68°C after 6 hours: Thermos's rated retention for a 0.5 L vacuum mug, which works out to just 0.043 W per °C.
- 535 times: how much more heat leaked into a NASA hydrogen tank's insulation with gas in it than in vacuum (3,845 W vs 7.18 W).
- 1.6%: share of perishable items still at a safe temperature in 705 packed lunches studied in Texas.
This is a collision: a steel dabba, a vacuum flask and an astronaut's suit should have nothing in common. They do. All three are solutions to one equation, the heat flow through a stack of thermal resistances in series, the same arithmetic as resistors in an electric circuit.
What changes across the three is which resistance does the heavy lifting. Let me solve each with real numbers, show where each one leaks, and see whether the claim in the title holds.
What are the three ways your lunch loses heat?
Heat leaves a hot dabba by conduction through solid walls, convection as air carries it away, and radiation as infrared light from every surface. Any insulator, from a cloth bag to a spacesuit, works by blocking one or more of these.
In 1892 the Scottish chemist James Dewar was trying to keep liquefied gases cold long enough to measure them. His fix was to put one brass vessel inside another and pump the air out of the gap, so heat could not cross by conduction or convection. Silvering the walls later cut the radiation too. Dewar never patented it. In 1904 two German glassblowers, Reinhold Burger and Albert Aschenbrenner, saw that the same flask could keep tea hot, built a sturdier version and named it Thermos. Dewar later went to court over the invention and lost. A device built for cryogenic physics became the lunch flask in millions of bags.
Wikipedia: Vacuum flask (history, citing Soulen, Physics Today, 1996)Each path has its own law, and naming them is half the solution:
- Fourier's law for conduction: heat flow through a wall is \( q = kA\,\Delta T/L \), so it falls with lower conductivity \(k\) and thicker walls \(L\).
- Newton's law of cooling for convection: \( q = hA\,\Delta T \), where \(h\) is low in still air and high in wind.
- Stefan–Boltzmann law for radiation: \( q = \varepsilon\sigma A (T_1^4 - T_2^4) \), where emissivity \(\varepsilon\) is near 1 for cloth and near 0.02 for polished silver.
Here is how the three containers handle each path:
| Path | Steel dabba in a bag | Vacuum flask | Spacesuit TMG |
|---|---|---|---|
| Conduction | Steel, k ≈ 17 W/m·K: no barrier | Only through the neck and lid | Thin spacer fabric between layers |
| Convection | Room air, slowed a little by the bag | None in the gap | None: space is a vacuum |
| Radiation | Polished steel ε ≈ 0.075 helps a bit | Silvered walls, ε ≈ 0.02–0.04 | Many aluminised films, ε ≈ 0.04 each |
| Main leak | Air at the outer surface | The lid | Seams, joints and compressed layers |
How fast does a plain steel dabba go cold?
By my estimate, with these example inputs, a 400 g lunch packed at 75°C in a plain steel dabba falls below 60°C in about 35 minutes in a 30°C room, and is near room temperature within 3 to 4 hours. Steel conducts so well that the only real resistance is the thin layer of still air outside it.
Assumptions
| Symbol | Meaning | Value | Basis |
|---|---|---|---|
| \(m\) | Food mass | 0.4 kg | Assumption: one dal-rice tier |
| \(c\) | Specific heat of moist cooked food | 3.6 kJ/kg·K | Estimate: mostly water (4.19) |
| \(A\) | Outer surface area | 0.05 m² | Estimate for a small two-tier dabba |
| \(h\) | Still-air convection plus radiation | 5.6 W/m²·K (range 3–10) | Assumption for still indoor air |
| \(T_0\), \(T_r\) | Packing and room temperature | 75°C, 30°C | Assumption: Indian summer office |
| \(T_s\) | Safe hot-holding line | 60°C | SFA, FSSAI training material |
Radiation from polished steel adds little to \(h\). Linearising Stefan–Boltzmann, \( h_{rad} \approx 4\varepsilon\sigma T^3 = 4 \times 0.075 \times 5.67\times10^{-8} \times 325^3 \approx 0.6 \) W/m²·K. A cotton bag with ε = 0.77 would radiate about 6.0 W/m²·K from its own surface, which is why the bag's air gap matters more than its fabric.
- Energy balance. The food's heat loss equals the flow out through the surface:
\[ mc\frac{dT}{dt} = -UA\,(T - T_r) \]
- Solve it. The temperature decays exponentially with time constant \(\tau = mc/UA\):
\[ T(t) = T_r + (T_0 - T_r)\,e^{-t/\tau} \]
- Time constant for bare steel. With \(U \approx h = 5.6\):
\[ \tau = \frac{0.4 \times 3600}{5.6 \times 0.05} = \frac{1440}{0.28} \approx 5{,}140\ \text{s} \approx 1.43\ \text{h} \]
Calculated in this post with these example inputs (0.4 kg, 0.05 m², h = 5.6, and a specific heat of 3.6 kJ/kg·K estimated from water at 4.19, Engineering ToolBox).
- Time to the 60°C line.
\[ t_{60} = \tau \ln\frac{T_0 - T_r}{T_s - T_r} = 1.43 \ln\frac{45}{30} \approx 0.58\ \text{h} \approx 35\ \text{min} \]
Researchers at the University of Texas measured food temperatures in 705 packed lunches from 235 families at nine child-care centres, using non-contact thermometers 1.5 hours before lunch. 39% of the lunches had no ice pack, 45.1% had one, and 88.2% were simply at room temperature. Of 1,361 perishable items, only 22 (1.6%) were in the safe range. Even lunches with several ice packs had more than 90% of items unsafe. The limit: these were cold lunches in Texas, measured once, not hot Indian food, but the physics of a thin container in still air is the same.
Read the paper (free)What does foam insulation add?
A layer of foam adds a large conduction resistance in series with the outside air, and 1 cm of polyurethane foam roughly triples the time a lunch stays hot. This is the insulated casserole or "hot case".
Resistances in series add, exactly like resistors:
Polyurethane foam has \(k \approx 0.03\) W/m·K, close to still air's 0.026, because foam is mostly trapped air in tiny cells too small to convect. With \(U = 1.95\), \(\tau = 1440/(1.95 \times 0.05) \approx 4.1\) h and the time to 60°C is about 100 minutes. Two centimetres of foam stretches it to about 165 minutes.
Calculated in this post from the foam conductivity of 0.03 W/m·K (Engineering ToolBox) and these example inputs.
Notice the trend: the foam's 0.333 already outweighs the air's 0.179. Once insulation dominates, the outside air barely matters, and the only way forward is a better insulator than still air. Still air is close to the floor for anything that contains gas, which is why the next step is to remove the gas.

Why does a vacuum beat any foam?
A vacuum removes conduction and convection across the gap completely, leaving only radiation, which silvered walls cut to a few watts per square metre. Thermos's own rating shows how good that is.
Thermos Japan rates its 0.5 L JNL-500 mug to stay above 68°C after 6 hours, starting from 95°C water in a 20°C room. Working backwards gives the flask's real heat-loss coefficient:
- Time constant from the rating.
\[ \tau = \frac{t}{\ln\frac{T_0 - T_r}{T - T_r}} = \frac{6\ \text{h}}{\ln\frac{95-20}{68-20}} = \frac{6}{0.446} \approx 13.4\ \text{h} \]
- Heat-loss coefficient. For 0.5 kg of water, \( mc = 0.5 \times 4186 = 2093 \) J/K:
\[ UA = \frac{mc}{\tau} = \frac{2093}{13.4 \times 3600} \approx 0.043\ \text{W/K} \]
Calculated in this post from the Thermos Japan rating and the specific heat of water, about 4,186 J/kg·K (Engineering ToolBox).
- Compare with the dabba. The bare steel dabba's \(UA\) was 0.28 W/K, about 6.5 times worse, for a similar size.
Calculated in this post from the Thermos Japan rating (0.043 W/K) and the dabba's 0.28 W/K, which uses these example inputs.
- Radiation across the gap. Between two parallel silvered walls with ε = 0.03, at 95°C and 20°C:
\[ q = \frac{\sigma (T_1^4 - T_2^4)}{\frac{1}{\varepsilon_1} + \frac{1}{\varepsilon_2} - 1} = \frac{5.67\times10^{-8}(368^4 - 293^4)}{65.7} \approx 9.5\ \text{W/m}^2 \]Over a gap area of about 0.04 m² (estimate), that is 0.38 W at a 75°C difference, or 0.005 W/K.
Calculated in this post from the emissivity of polished silver, 0.02 to 0.03 (Engineering ToolBox), the Stefan–Boltzmann constant, the Thermos Japan rating temperatures (95°C and 20°C) and an estimated gap area of 0.04 m².
So radiation through the walls explains only about an eighth of the measured 0.043 W/K. The rest is the lid and neck, where solid plastic and steel bridge the vacuum. That is a real design lesson: fill a vacuum jar to the top, keep the lid shut, and pre-heat it with hot water so the inner wall isn't stealing heat from the food.
See the data as a table
| Bare steel dabba (°C) | 1 cm foam casserole (°C) | Vacuum jar (Thermos-class) (°C) | |
|---|---|---|---|
| 0 h | 75 | 75 | 75 |
| 0.5 h | 61.7 | 69.8 | 72.6 |
| 1 h | 52.3 | 65.3 | 70.4 |
| 2 h | 41.1 | 57.6 | 66.3 |
| 3 h | 35.5 | 51.6 | 62.5 |
| 4 h | 32.7 | 46.9 | 59.2 |
| 5 h | 31.4 | 43.3 | 56.2 |
| 6 h | 30.7 | 40.4 | 53.5 |
Source: Newton-cooling model (this post); vacuum curve uses UA = 0.043 W/K derived from Thermos Japan JNL-500 rating · Estimates. Assumes h = 5.6 W/m²K, A = 0.05 m², c = 3.6 kJ/kg·K, foam k = 0.03 W/m·K. Hot food is safe above 60°C.
How does a spacesuit use the same trick?
A spacesuit's outer garment is a vacuum flask turned inside out: space supplies the vacuum, and many layers of aluminised film stop radiation one reflection at a time. This is multilayer insulation (MLI), and it is used on almost every spacecraft.
Without it, a suit surface at body temperature with ordinary fabric emissivity would radiate into deep space at
On roughly 2 m² of suit (my estimate) that is about 780 W, several times the heat a resting person makes. In shadow the astronaut would chill fast.
The radiation-shield equation
Each shiny layer acts as a radiation shield. With \(N\) shields of equal emissivity between a hot wall \(T_h\) and a cold wall \(T_c\), the textbook result is:
- Seven shields, ε = 0.04, 300 K inside, 100 K outer layer (lunar-night scale).
\[ q_7 = \frac{5.67\times10^{-8}(300^4 - 100^4)}{8 \times 49} = \frac{454}{392} \approx 1.16\ \text{W/m}^2 \]
- Fifteen shields. \( q_{15} \approx 454/(16 \times 49) \approx 0.58 \) W/m², half again.
- Compare. From 390 W/m² bare to about 1 W/m²: a cut of roughly 300 to 670 times.
Calculated in this post from the two results above, with ε = 0.04 for aluminised film (aluminium foil, Engineering ToolBox) and an assumed ε = 0.85 for ordinary suit fabric (cotton cloth is 0.77 there).
The shuttle-era suit's thermal micrometeoroid garment used a liner of aluminised Mylar layers under an Ortho Fabric shell; secondary sources give about seven such layers, so treat the exact count as approximate.
NASA wrapped a 1.39 m spherical liquid-hydrogen tank in two MLI blankets, each with 15 double-aluminised Mylar shields separated by 16 silk-net spacers, and ran 19 thermal cycles. In vacuum ("space-hold") the heat leak was 7.18 W; with helium gas filling the blankets on the ground it was 3,845 W, about 535 times more. Moisture raised the vacuum heat leak by up to 19%, to 8.91 W. The limit: a cryogenic tank at 20 K, not a suit at body temperature, but it measures exactly the effect of removing gas from layered insulation.
Read the report (free)The NASA number also shows where the ideal equation flatters reality. 7.18 W over the tank's 6.07 m² is 1.18 W/m². The ideal formula for 30 shields at ε = 0.03 between 300 K and 20 K gives only 0.23 W/m². Real blankets do about five times worse than ideal because of spacer conduction, seams and compression, and a bending, crumpling spacesuit is worse still.

A NASA Johnson Space Center team reviews the outer shell fabrics of spacesuits and a project to replace them for the Moon. The current EMU, first delivered in 1982, is protected by a thermal micrometeoroid garment with an Ortho Fabric shell rated from about −184°C to 149°C; its white Gore-Tex face reflects most sunlight. Artemis suits must survive 127°C to −173°C for 8 hours, and −243°C (30 K) in permanently shadowed craters for up to 2 hours. The limit: it focuses on the outer fabric's durability and dust, not on the full heat balance of the suit.
Read the paper (free)What if the assumptions change?
The ranking of the three never changes, but the minutes do. Here is a sensitivity check on the steel dabba, the case with the most guesswork:
- Still air, \(h\) = 3: τ = 2.7 h, so about 65 minutes to 60°C.
Calculated in this post from these example inputs with one value changed, and the 60°C hot-holding line (Singapore Food Agency).
- A fan or breeze, \(h\) = 10: τ = 0.8 h, so about 19 minutes.
Calculated in this post from these example inputs with one value changed, and the 60°C hot-holding line (Singapore Food Agency).
- A 25°C air-conditioned office: the gap to room grows, and the time to 60°C falls to about 31 minutes.
Calculated in this post from these example inputs with one value changed, and the 60°C hot-holding line (Singapore Food Agency).
- Twice the food (800 g): the time to 60°C doubles to about 70 minutes, because mass grows faster than area. This is the square-cube law, and it is why a full three-tier dabba holds heat better than one small box.
Calculated in this post from these example inputs with one value changed, and the 60°C hot-holding line (Singapore Food Agency).
For the vacuum jar, the result rests on Thermos's measured rating, so the main uncertainty is food versus water: food's lower specific heat shortens the time by about 14%.
Calculated in this post from the specific heat of food, 3.6 kJ/kg·K (an estimate), against 4.19 for water (Engineering ToolBox).
Where does the analogy break down?
The physics is shared; the job is not. Three honest differences:
- The suit's real problem is getting rid of heat. A working astronaut makes more heat than the insulated suit can leak, so suits carry a water-cooled garment and an active cooling system. A dabba only ever loses heat.
- The vacuum is free in space. A flask must build and hold its own vacuum; a suit gets it from outside, which is why MLI works in orbit but would be mediocre in a lunch bag full of air.
- Sunlight. Radiation flows into a suit as well as out. Its white outer fabric reflects sunlight, a job no dabba has to do.
Check your own lunch
Use this to check what you just read. Set your food, your packing temperature and your insulation, and see how long it stays above 60°C.
How long will my lunch stay hot?
A dabba, a thermos and a spacesuit really are one design at three scales: each is a stack of thermal resistances against conduction, convection and radiation. A bare steel dabba has a heat-loss coefficient of about 0.28 W/K and keeps lunch above 60°C, with these example inputs, for roughly 35 minutes; a vacuum jar's 0.043 W/K stretches that to nearly 4 hours; and a suit's layered film in vacuum cuts radiation from about 390 W/m² to about 1 W/m². The analogy breaks only on purpose: the suit mostly needs to shed heat, and the flask must make its own vacuum.
Calculated in this post from the Thermos Japan rating (0.043 W/K), the example inputs above, and the emissivities in the Engineering ToolBox table (Engineering ToolBox).
Insulation is never about adding warmth. It is about making every path out of the box a little longer.

More on keeping food safe and hot: food on long motorcycle rides in India, whether dhaba food is safe on the highway, the science of the pressure cooker and how Bengaluru's darshinis serve food hot and fast. The same layering logic is in riding gear, collected on the gear page.
What's new
- NASA's Artemis III moonwalk suit passes 850 hours of crewed pressurised testing NASA, Feb 2026
- Axiom's AxEMU suit completes its first uncrewed thermal vacuum test Axiom Space, Nov 2025
- NASA team sets out new outer fabrics for lunar suits ICES, Jul 2025
What if next?
- Tadka is a chemistry extraction, and so is perfume: what hot oil does with heat instead of keeping it.
- Idli batter and bread dough are the same factory: where you want food to stay warm, for the microbes.
- A pani puri is an engineering problem: another thin shell racing a clock.
Questions people ask
How long does food stay hot in a steel tiffin?
Not long. By a Newton-cooling estimate with these example inputs, 400 g of food packed at 75°C in a plain steel dabba falls below the 60°C hot-holding line in about 35 minutes in a 30°C room, and nears room temperature in 3 to 4 hours.
Calculated in this post with these example inputs and the 60°C hot-holding line (Singapore Food Agency).
Is a vacuum flask better than an insulated casserole?
Yes, by about 1.5 to 2 times by my estimate with these example inputs (225 minutes against 100 to 165 for 1 to 2 cm of foam). Foam is mostly trapped air, so it can never beat still air's conductivity of about 0.026 W/m·K. A vacuum removes conduction and convection entirely, leaving only radiation and the lid.
Why are thermos flasks silvered inside?
To cut radiation. Polished silver has an emissivity of about 0.02 to 0.03, against about 0.96 for paint, so the shiny walls radiate and absorb far less infrared across the vacuum gap.
How do spacesuits keep astronauts warm?
The outer garment uses multilayer insulation: several aluminised plastic films that reflect infrared, working in the vacuum of space. Suits also carry an active cooling system, because a working astronaut makes more heat than the insulation lets out.
Sources
- Almansour FD et al.: Temperature of foods sent by parents of preschool-aged children, Pediatrics 128(3):519–523 (2011) (705 lunches, 1.6% of perishables safe)
- Sumner IE: Thermal performance of a helium-purged tank-mounted multilayer insulation system, NASA TP-1114 (1978) (7.18 W in vacuum vs 3,845 W with gas; 15 double-aluminised Mylar shields per blanket)
- Jones RJ et al.: Artemis Suit Material Project Overview, ICES-2025-56 (2025) (EMU from 1982, Ortho Fabric −184 to 149°C, Artemis 127 to −173°C)
- Thermos Japan: JNL-500 vacuum insulated mobile mug specifications (above 68°C after 6 h from 95°C at 20°C)
- Engineering ToolBox: Thermal conductivity of common materials (air 0.026, PU foam 0.03, stainless steel 17 W/m·K)
- Engineering ToolBox: Emissivity coefficients (polished silver 0.02–0.03, polished stainless 0.075, cotton 0.77)
- Singapore Food Agency: Proper temperature for hot foods (above 60°C; danger zone 5–60°C; eat within 4 hours)
- Wikipedia: Vacuum flask (Dewar 1892, Thermos 1904, Dewar's lost court case)
- NASA: Moon mission spacesuit nears milestone (2026) (850 hours of crewed testing)
- Axiom Space: First uncrewed thermal vacuum test of AxEMU (2025)
