Why are early submarines faster on the surface, while later types are faster when submerged?

In the vein of Mark's more or less apt characterization of PhillS's reasoning (" ...but approaches it backwards... "), I would say his answer needs a slight adjustment of point of view.

Contrary to Alex's " ...is larger because of the hydrodynamic reasons..., the main thing limiting a submarine's (or any ship's) speed, is the power and duration-capability of its drive. That applies above or below the surface. (Although, an argument could be made that a submerged surface ship...)

When driven by a man-powered screw, submarines were especially non-fast. Diesel engines made them faster, but fuel consumption if going, say, 20 knots would severely limit mission availability versus going perhaps 11-12 knots to one's station and transiting about once there. Fuel consumption rises much faster than speed.

But diesel engines only work underwater when running, using oxygen and fuel, and producing smoke. No one ever found a terrific way to store the smoke onboard or to hide it if exhausted, and while snorkeling both brought air in and got rid of smoke, the smoke was visible in one way or another. Engineers even thought of carrying hydrogen peroxide as a source of oxygen (in a volume as large as the ship's "people tank" would have been), and thereby running the diesels without surfacing and at a depth that would greatly lower the still exhausted smoke's impact on the surface of the water and in the air, but it wasn't a... winner of an idea.

Nuclear power, on the other hand, had none of those difficulties. It introduced its own issues vis-a-vis detectability (pump noise, for example), but while those things were continuous while a set of running diesel generators were intermittant, they are in NO way as detectable as a set of running diesel generators.

And the final nail for diesel-powered submarines as to speed is size. Their power scales up (along with fuel use and it's required storage space, which rises with ship speed exponentially, not multiplicatively), but it is a losing battle.

So the compromise was mediocre speed transiting to and on station and in short chases powered by the diesel engines and a hull shape to amplify that set of conditions.

Actual attack, generally, but not always, involving a submerged approach, was done with the much less detectable electrical-powered engine. Electricity from batteries. And those batteries were both limited in size by the technology of the day AND by their explosion hazard. The latter was a bigger issue 80 years ago than now, but even today, controlling current flow from and to the battery to keep its hydrogen production at safe levels (a multiple variable subject) is of HUGE importance. And batteries are large. Think a hundred cells each four feet tall. To provide a sufficient current to go that (mediocre, "tramp freighter" class) 11-12 knot speed would have been to big to carry, maintain, or vent the hydrogen produced. Not to mention creating the distilled water to keep replacing water being split into gasses.

So the requirement to limit one's attack or fleeing from attack movement to being electrically powered limited their speed underwater versus their (mediocre) speed on the surface.

Now herald nuclear power. Fuel is, in a non-strategic sense, unlimited. One does not have to use electrical power to drive the ship. The steam plant drives it (sorry PhillS, steam, not electricity, drives the main engine shaft/propeller, not electricity) in all conditions one cares to use it. One may NOT care to use it, more in a second, and it is able to power the shaft at 100% of its capability, surfaced or submerged. So the ship can always go its top speed for either situation, at all times, always. At that point your engineering choices push down a level to things like hull shape and what weapons to use. No mysteries here. Since the ship is vastly harder to even detect, much less fight against, when underwater, one can decide to keep it underwater at all times, and decide accordingly.

Missions also change. The anti-shipping mission was given to aircraft. Hiding nuclear missiles, or sinking Soviet ships, became the missions (though I suppose the Soviets would have said "sinking NATO ships") and the former required, the latter demanded, that submerged sailing dominate one's choices. So the hull and appurtenances were designed to best support that.

Since the power available is the same, a nuclear sub could be designed to travel best on the surface, but even if the modern missions did not need submerged sailing, one would wonder WHY one would want a sub to spend a lot of its life on the surface as there are already ship types that can do anything it is able to do hugely better than a surfaced sub ever could.

So, the reason why they are faster underwater is that since we made nuclear power an option, every design consideration has been made with that mode as the ONLY mode of interest. (With the understood exception of how one is to enter and exit the ship.)

An aside, since the absolutely large interest around the world in diesel submarines has been mentioned, one might point out two things:

1. There is no nuclear-powered submarine that can be quieter than a modern electrical-powered submarine. Bearing in mind that nukes CAN and do run off their batteries that does not mean they have a weakness here, but for nations that care to spend their money differently, or that refuse to use nuclear power, a modern diesel-powered submarine is absolutely a wonderful thing for coastal/national waters missions. Few of these nations are trying to project power ten thousand miles and for months on end. But they are still slower under water than above it. And so is a nuclear sub when running on its electrical power.

2. It's still all about the power you can deliver, nothing else. All the rest becomes choices. Nuclear power plants are amazing in an unexpected way: they overwhelm the situation. Usually, folks make dollar-based decisions and so one would never expect something that so overwhelms the most one can even unreasonably require. Different main engine, in whatever ways one would need to change it, and a submarine could go as fast as an aircraft carrier while submerged. And on the surface. Subs have gotten much bigger over the decades, but still never outstripped the power plants available. There is simply no condition in normal, non-damaged operation in which the ship's power plant can't provide flank speed. Well, I never experienced 100% power when running flat out at the max shaft revs allowed. You just don't see that in the engineering world too often. And still quiet enough to position oneself to attack 30+ knot Soviet cruisers racing out of the bottle to open sea (missile sub, in the fantasy that we'd've survived a missile launch when the Soviet backshot would reach us before or right about when the last missile bumped out of its tube, not an attack sub, so we weren't chasing those cruisers down, just placing ourselves and taking shots... if war had happened). Much better positioning than a 3-4 knot electric engine driven diesel sub and all due to our nuclear power plant.

You are comparing diesel submarines with nuclear ones. Diesel submarines are slower under water because of the very limited capacity of accumulators, especially in the submarines of WWII times. Nuclear submarines are powered by the same engine when submerged and when on the surface, so their speed when submerged is larger because of the hydrodynamic reasons: a submerges submarine almost does not create a surface wave, and the surface wave is the main reason which limits the speed of a surface ship.

It's not correct that all early designs were faster on the surface rather than submerged. (Nor was it always true that being submerged offered better stealth than being surfaced.)

The focus on engine type in the other answers is very relevant to the duration a submarine can spend submerged, but not so much to speed. (Consider, for example, that a Tesla electric car is not inherently slower than a gasoline-powered vehicle just because it's using an electric engine.)

The primary determinant of whether a a submarine is faster on the surface or below it is the hull shape (along with control surface design and propeller location): given similar engine power a "boat" style hull is faster on the surface and a teardrop hull is faster when submerged. (The teardrop hull is slower on the surface for various reasons, including how it bow wave acts as compared to a boat hull.)

Note that on submarines the teardrop hull is not significantly newer than the boat-style hull; a number of early submarines, such as the 1888 Peral, used the teardrop hull and single central propeller preceded by cruciform control surfaces bias towards good underwater performance over good surface performance:

(For more examples of early teardrop hulls, see this chronological series of images.)

Now in submarines up to about the middle of World War II, it's true that propulsion technology had a significant effect on whether submarines were designed to run faster on the surface or underwater, but much of this was to do with general operational doctrine (i.e., where and how the submarines were designed to be used) and one of the key issues was not speed on versus below the surface but submarine speed period. Early diesel engines could produce nowhere near the power that steam turbines could produce, and in doctrines where submarines were expected to try to keep up with surface fleet ships the designers were stuck behind surface ships both underwater and on the surface. This led to such, uh, "interesting" designs such as the 1913 steam-turbine powered British K class (24 knots surfaced), which unlike the diesels could match battleship speeds.

Yet even at this point, well before World War II, there were submarines designed for different doctrines that ran faster submerged than surfaced. Examples range from the 1917 British R class (9.5 knots surfaced, 14 knots submerged) to the 1938 Japanese No. 71 (13.25 knots surfaced, 21.25 knots submerged), albeit some of these, such as the No. 71, were experimental.

By the start of World War II, doctrine was coming around to the idea of submarines operating independently against merchant shipping, rather than being part of a battle fleet or even just operating against individual cruisers or the like. One of the reasons for this was that submarines still were nowhere near keeping up with battle fleets (steam turbines fired by coal or oil turned out, unsurprisingly, not to be a very good idea in submarines); if anything they were falling further behind as battleship speeds had by this point increased from 20-24 knots to well over 30 knots.

For this doctrine, a key need was (relatively) long range: submarines had to do a lot of cruising to find merchant ships in the vast Atlantic (or even vaster Pacific): diesel engines were the only practical option for this at the time. With sonar at this point significantly better than radar (meaning you're less detectable at night when surfaced), and the snorkel yet to be invented anyway, tuning for surface range was the obvious way to go. Note that this is similar to, but not the same as, tuning for surface speed; it's best to think of the better surface speed as a side effect of the tuning for range.

As radar was introduced and improved, it became better to be submerged rather than surfaced, and the snorkel enabled this. Nothing could be done about the hull shape of existing boats, of course, but new designs starting to be built around mid-war (e.g. the 1942 German Type XVII and 1943 German Type XXI) were now using streamlined hulls more optimised for submerged running speed and running faster submerged than surfaced. Note that this was a change primarily driven by doctrine: the propulsion technology of the Type XXI was not a huge advance over the later Type IX boats, and in fact the Type XXI was almost three knots slower on the surface than the Type IX, despite having improved diesel engines. Post-war, the Type XXI's return to hulls optimised for underwater speed went on to influence submarine design world-wide, including the Soviet Whiskey, American Tang, British Porpoise, Swedish Hajen III and many others.

Interestingly enough, just after WWII the Type XXI hull improvements were also applied to some existing American "boat hull" submarines via the GUPPY programme. This did involve some propulsion changes, including a higher-capacity, higher-current battery system and new motors, but the primary change was to the hull design (a completely new bow and extensive streamlining). This changed the performance of the Tench-class Pomodon from 20.25/8.75 knots surfaced/submerged to 17.8/18.2 knots. While it may not be so obvious that the huge underwater performance increase was due more to the hull than the electric propulsion changes, the reduction in surface speed on the same diesel engines is obvious.

Post-WWII the advent of nuclear power was a huge technology change, of course (finally steam turbines could be used!), but this did not automatically translate to submerged speed being faster than surfaced speed. As always; it depended on the hull design. Even the United State's second nuclear design, the USS Seawolf ran faster surfaced (23 knots) than submerged (19 knots).

And looking solely at the advent of nuclear power similarly fails to explain the speed reversal in conventional diesel-electric submarines, which even into the '80s were using underwater propulsion technology not hugely better than WWII. The highly popular 1971 German Type 209 is a good example: it does 11.5 knots surfaced, 22.5 knots submerged, still though with very limited underwater range on electrics; going to (or back to, from an 1880s point of view) a teardrop hull design is by far the most significant factor here. Again, it comes down more to doctrine—how the submarines are intended to be used—not merely what technologies are available.

Some of this is a history question, and some isn't.

Firstly the non-history part. All things being equal, submarines should move faster underwater. Your intuition that there is more drag underwater is incorrect. True, there is more drag from the effect of fluid flowing over the surface of the submarine, but on the surface the effect of drag from the surface itself is more important. Surface effect (surface tension) and the ability to produce and interact with waves on the surface more than offset the fact that the submerged surface is less.

This is why for example competitive swimmers try to stay underwater as long as the rules allow after a turn in racing - gliding in a streamlined position completely underwater is faster than being on the surface and actively propelling themselves. So this explains why nuclear submarines have higher speeds underwater.

Why were WWII era submarines faster on the surface? That's the history bit. They weren't nuclear powered - they had diesel engines, which only work on the surface. (You really don't want to be using up oxygen and producing diesel fumes when you are in an enclosed environment underwater). They were essentially surface ships, powered by a diesel engine. But, they had the capacity to dive underwater, and had electric motors (powered by batteries) for when they were submerged. These motors produced considerably less power than the diesel engines and couldn't propel the submarine as fast.

For comparison, nuclear submarines use a nuclear reactor to generate electricity and drive the submarines electric motors, and it uses the same system on the surface and underwater, since it doesn't have the problem of limited battery life. And nuclear reactors don't have the same problem of producing exhaust fumes and using up oxygen that prevented diesel engines from being used when submerged.