Drywell (D/W)

The D/W, shaped like an inverted lightbulb, and the suppression chamber (S/C), compose the CV. The D/W and S/C are connected with eight thick vent pipes. When gas travels to the S/C from the D/W, it bubbles up in the S/C pool. The name drywell comes from the absence of water in contrast to the S/C, also called the wet well.

Suppression chamber (S/C)

The S/C is the toroidal vessel that makes the lower part of CV. The Unit 1 S/C contains as much as 1,750 tons of water and those of Units 2 through 4, 2,980 tons. After accidents like a pipe rupture or SRV activation, high-temperature steam discharges into the S/C, and the cooled steam condenses to water to control the CV pressure from rising (Figure 2.2). S/C is also called the pressure S/C. The chamber also supplies water to emergency systems like reactor core isolation cooling system (RCIC) or high-pressure coolant injection system (HPCI) and is also called the wet well or torus.

Safety relief valve (SRV)

Safety relief valves activate to reduce the RPV pressure when it exceeds the allowable value (see Section 2.11.3 for details). Each reactor has eight SRVs (four in the case of Unit 1), which share a number of functions among them.

Vent valve

Vent valves open to vent the steam inside the CV to the outside in case of excessive CV pressure during a severe accident. The vent line has two different types of valves that both had to open for venting.

Diesel-driven fire pump (D/DFP)

D/DFP are fire pumps driven by diesel engines. Each reactor has one unit to back up the electrically driven fire pump.

Metal-clad (M/C) switch gear

This is a switchboard for high voltage of 6900 V. The power drives large-size equipment like the seawater pumps and condensate pumps. M/C, as Figure 2.3 shows, is the primary equipment of the electrical power systems.

Power center (P/C)

P/C is the switchboard for distributing AC low-power voltage of 480 V. Many machines throughout a power plant are driven with power from this switchboard.

Direct current (D/C)

D/C systems transform the AC power from the P/C to D/C and charge batteries as needed. D/C supplies power for control and instrumentation and driving D/C motor-operated (M/O) valves and is the most important power source for power plants. D/C, backed up by batteries, was believed to have low risk in its loss; however, this accident destroyed many of them. There are three types of D/C: 250, 125, and 24 V.

Core cooling systems during normal operation

Steam with high temperature and high pressure from the nuclear reactor turns the turbine, is routed to the main condenser, and returns to the reactor as water. Nuclear fission in the reactor core during normal operation generates a large amount of heat. About one-third of the heat turns into electricity, and the main condenser has to remove the remaining two-thirds so it has a large cooling capacity. The heat is then removed at the heat exchanger and discarded into the sea.

Core cooling systems during normal shutdown (including after emergency shutdown [SCRAM])

Nuclear fission stops after normal shutdown for periodic maintenance or emergency shutdown following causes like an earthquake. The core, however, continues to generate decay heat. Without cooling, the reactor has no place to discard its heat, so the residual heat removal system (RHR), although with a capacity smaller than the main condenser, starts running to cool the core.

Emergency cooling systems

Normal shutdown is a condition of normal operation. SCRAM that might be caused by an earthquake is also a normal operation designed to take place in such emergency situations. For preparation against abnormal emergency events like RHR inoperability, which was experienced with this earthquake, or loss of coolant accident (LOCA) (see Remarks) from a piping failure, the reactor is equipped with a number of emergency cooling systems. Some may require electricity for their operation, but others like isolation condenser (IC) with Unit 1 and RCIC with Units 2-4 work without electrical power. High-pressure systems can inject water even when the reactor is at high normal operation pressure of about 7 MPa, whereas the lower pressure systems can only do so when the reactor pressure is at 1 MPa or less. These emergency cooling systems offer a variety to counter various situations.

Isolation condenser (IC)

lC is a condenser for the emergency situation of core isolation. This type of core cooling system was only present with Unit 1. It operates even with high reactor core pressure with natural convection without any power sources (Figure 2.5). The system can withstand an extended operation by supplying cooling water to the condensate tank.

When the earthquake hit, Units 1 through 3 were in normal operation, and 4 through 6 were in scheduled shutdown. Upon the earthquake attack, Units 1 through 3 automatically SCRAMed, and the operators switched the electric power to the site to an external power source. Almost at the same time of the switching, external power was lost (Figures 2.6 and 2.7), and the emergency D/Gs (Figure 2.8) kicked in automatically. As part of the fail-sale function, the main steam isolation valves (MSIVs, valves to close the main steam line to the steam turbine) automatically closed. The series of processes following the SCRAM proceeded normally.

March 11, about 14:50: Unit 2 RCIC started manually

With the SCRAM, the feedwater pump for Unit 2 stopped, and following the manual, the operator manually started the RCIC with remote operation (turning the switch in the central control room). The RCIC, however, after about a minute of running, automatically stopped because it detected a high water level in the nuclear reactor. The high water level was due to the injection capacity of 95 ton/h with the RCIC, which had started injecting water from the external condensate storage tank to the RPV. Eleven minutes later, the operator restarted the RCIC remotely with his eyes on the reactor water level.

March 11, 2011, 14:52: Unit 1 IC started automatically

The MSIV closure with Unit 1 resulted in a reactor pressure rise, which triggered the IC to automatically start. At about 15:03, however, the operator, worried about the rapid reactor temperature drop (see Remarks), remotely stopped the IC. From that time until 15:30, the operator started and stopped the IC three times, controlling the reactor pressure at 6-7 MPa.

March 11, 15:05: Unit 3 RCIC started manually

In the same manner as with Unit 2 the operator started the Unit 3 RCIC remotely. At about 15:25, however, on detecting high reactor water level, the RCIC stopped automatically. At this time, the reactor pressure values with Units 2 and 3 were too high, and the pressure relief functions of the SRVs had already started to discharge RPV steam into the S/C.

March 11, about 15:27: first tsunami wave

The first tsunami wave arrived; however, it was low enough to not cause damage that led to the SBO.

March 11, about 15:35: second tsunami wave

The second tsunami wave reached the coast. The inundation height reached 11.5-15 m and even 17 m at some places. First, the seawater pumps for emergency D/Gs, located near the coastline at an elevation of 4 m above sea level, suffered water damage (Figure 2.9). The details of the damage are unknown, but they lost all functions.

The plant main area with the reactor building and turbine building was at 10 m above sea level; thus, they suffered a maximum inundation of 7 m. The water entered the buildings through doors and air ventilation holes, and most facilities in the basement of the turbine building, including the switchboards, were submerged. This wave caused the loss of almost all electric power sources and was the starting point of this severe accident. The plant response headquarters were set inside the seismic isolation building in a room with no windows and no cameras to look outside; thus, for a time after the SBO, headquarters staff could not tell that the problems were caused by the tsunami.

One report after another reached headquarters with the news of loss of AC power with all units and even DC power with Units 1, 2, and 4. Headquarters staff was stunned by the scope of a catastrophe beyond any imagined scenarios. The plant manager realized that the situation was beyond any hypothesized severe accidents, but could not decide what immediate action to take. He thus followed regulations and reported the SBO to the ministers and agents.

The states of water submergence were about the same for Units 1 through 3; seawater entered the turbine buildings from the equipment hatch and ventilation holes to flood the basement and the concourse floor between the basement and first floor of the turbine, control, and reactor buildings. The basement of the turbine buildings contained the emergency generators and the regular and emergency switchboards; the control building basements contained the DC power systems (Units 1, 2, and 4); and the reactor building basements were where the pumps and other equipment for the emergency cooling systems of RCIC and HPCI were staged (Figure 2.10).

When the seawater pumps fail, the nuclear reactor loses its heat sink (where to dump excessive heat) to the ocean. The Unit 1 IC was designed to release steam to the atmosphere. Units 2 and 3 with RCIC could transfer the heat to the S/C; however, they had lost their ultimate heat sinks. The Unit 1 IC also failed to function; thus, all units that were up and running at the time of the accident had lost their ultimate heat sinks other than venting the CV to release the heat into the atmosphere.

March 11, about 15:39: Unit 2 RCIC started manually just before the tsunami damage

At about this time, an operator opened the Unit 2 RCIC valves remotely. This action was immediately followed by the SBO. If the opening had been delayed by seconds, the Unit 2 RCIC would have been lost (see Remarks), and situations with Unit 2 would probably have been worse than the actual trace of events. The runaway Unit 2 RCIC kept running for about 3 days.

March 11, 15:37-42: loss of all AC power

Except for the Unit 6 air-cooled emergency D/G (6B), all AC power sources for Units 1 through 6 were lost. Units 1 and 2 lost their DC power as well. Unit 3 had its DC power on the concourse level between the basement and first floor. The unit was covered with water, but managed to retain its function; Unit 3 allowed operation of its RCIC and HPCI for a few days.

Unit 1 lost all its electrical power including DC. The central control room was pitch dark, and the loss of DC power stopped displays of all instruments. The most critical reactor water level and pressure became unknown.

The isolation condenser (IC) was in normal operation with the operator turning it on and off. The loss of DC power, however, caused the fail-safe logic to send “close” signals to all four valves. High-temperature steam from the reactor stopped feeding to the IC to be condensed into water to run back to the reactor. The cooling function of the IC was almost completely lost. The Unit 1 IC was designed that on failure (an abnormal condition), all flows were stopped to prevent possible radioactivity leakage and to contain the pressure vessel. This concept, however, had the contradiction that shutting off the IC cooling function could lead to a more dangerous situation.

Around this time, Plant Manager Yoshida was not sure whether the Unit 1 IC and Units 2 and 3 RCIC were properly operating, but at the same time was not aware of the Unit 1 IC stoppage with the fail-safe function.

Unit 1 lost both the DC for control and AC for driving power, and when the “valve close” signals were sent, the driving power to make those actions (480 Vac for the two valves inside the CV and 125 Vdc for the two outside) were also lost. Thus, whether the valves actually closed or stopped midway is unclear. The fact of rise in radioactivity recorded only 2 h after the SBO at about 18:00, and other evidences confirm that the IC was not functioning after the SBO (see Section 2.5.3 for details).

Another root cause for failing to recognize the IC stoppage, in addition to the preceding design problem, was the lack of education and training for the hypothetical situation of loss of DC power. Furthermore, for this March 11 accident and the hours that followed, none of the people that were responding to the accident, in the plant response headquarters, main office response headquarters, Nuclear and Industrial Safety Agency (NISA), or Nuclear Safety Commission (NSC), recognized the IC stoppage. One of the reasons why the operator on duty did not recognize the fact was because no one had ever experienced the IC in operation. The best knowledge that anyone had was having heard from former operators that when IC was in operation, steam from the condenser cooling water would blow out horizontally from the exhaust openings (the so-called pig snout) to produce lightning-like static and large thunder-like sounds (Figures 2.15 and 2.16). The reason why no one had experience of the event was because Unit 1 was the oldest among those with MARK-I containment. There was only one other reactor, Tsuruga-1 of The Japan Atomic Power Company (JAPC), that had adopted IC for its cooling system in the event of reactor core isolation.

The Unit 1 operator was puzzled by the small amount of steam from the IC condenser, but did not have a good understanding of the situation. He thus switched Valve 3A from “Open” to “Close” and “Open” again three times between 18:00 and past 21:00. Because the three valves other than 3A remained (almost) shut due to the fail-safe function, operating Valve 3A was meaningless. As a result, Unit 1 was in an abnormal situation with no water injected into the core until water injection from a fire engine started at 4:00 the next morning. This is how the core melting progressed rapidly with Unit 1, for which water injection had stopped at only 1 h after its SCRAM when the decay heat was still high.

March 11, 16:42: drop in water level detected

At about this time, operators recognized that the indicator of the wide range water level monitor (one of the three types of reactor water level monitors) was working, but the reason for this sudden recovery remains unknown. The indicator at first was pointing at “−900 mm,” and 14 min later after showing “−1500 mm,” it went off scale. During this time period, the water level monitor was probably working normally, and the plant response headquarters, given information about the rapid drop in water level, estimated the time for the water level to drop to top of active fuel (TAF) to be an hour; however, in the midst of all confusion, headquarters did not recognize the IC stoppage.

March 11, 17:12: use of water injection from fire engines evaluated

Under SBO, the AM procedure designates the only alternative method for water injection other than IC to be D/DFP driven water injection. The water source in that case was the “filtered water tank (see Remarks 1).” Plant Manager Yoshida was worried about piping from the filtered water tank to the reactor building. He then ordered evaluation of water injection from a fire engine, although it was not prepared as an AM procedure. On March 12th and later, this water injection from fire engines turned out to be the lifeline. At this time there were three fire engines on-site; however, one was unusable with damage from the tsunami, and another was parked near Unit 6, with its way to Unit 1 blocked by tsunami rubble. That left only one engine available for the purpose (see Remarks 2). To make things worse, alternative water injection with a fire engine was not written in the AM procedure, and no team was assigned to take responsibility for the action. The plant manager had given orders but no progress was made until before dawn on the next day, March 12th.

March 11, 17:30: doubts about the operation of IC

Some operators on duty in the central control room had doubts about the operation of IC. In preparation for alternative water injection, the operator checked the D/DFP start-up and put it on standby. Later, at 20:50, he started the D/DFP engine to prepare for the change of situation.

Later in the afternoon, four 6 V batteries from a neighboring company and two 12 V batteries from large buses arrived. They were hooked up to the back of the control panel in the central control room to start monitoring water levels of Units 1 and 2.

March 11, 17:50: first abnormal radioactivity detected

A worker approached the reactor building to check the water level of the IC condensate tank; however, near the double door (see Remarks), radiation dosage read as high as 2.5 μSv/h, and he had to retreat. This high-dose reading was much smaller compared to later large readings. However, it was out of normal range, and probably at this time, part of the fuel was exposed above the top of cooling water.

March 11, 18:18: IC valves 2A and 3A closed

In the central control room, the operators realized the IC lamps that had been off were illuminated showing “Valve 2A and 3A Closed.” At this time the operator on duty thought that all IC valves might be closed with the fail-safe function, but hoping they were not, he remotely opened valves 2A and 3A. He then reported to the response headquarters about the opening. The operator left the central control room to visually check the steam ejection from the reactor building. The operator only saw a small amount of steam at first, and then no more steam followed the initial exhaust.

March 11, 18:25: IC valve 3A remotely closed

Seeing the small amount of steam from the IC, the operator who had doubts about the IC function began to worry that “steam not blowing out from the IC may be a sign of low condenser cooling water level, and keep running the IC may cause damage to the IC piping,” [9,10] and at 18:25 he remotely closed Valve 3A. This operation, however, was not reported to the response headquarters.

March 11, 20:07: RPV pressure acknowledged

A worker entered the reactor building and found the RPV pressure at 6.9 MPa. Around this time, the SRV might had been open with its safety function because the Unit 1 SRVs were set to open at 7.7 MPa and remain open until the RPV pressure drops to about 6.9 MPa. Up to this time, however, workers were allowed in the torus in the basement of the reactor building; thus, a large amount of radioactivity release through the SRVs to the S/C had been unlikely. Within the reactor, therefore, a certain amount of fuel damage at the time cannot be denied; however, it is natural to assume that the core had not reached significant core damage to release a large amount of radioactivity.

March 11, 21:19: water level gauges started to give misreadings

The water level gauge again read “TAF + 450 mm.” The water level at this time, however, was unlikely to be above the TAF, so it is most likely that from this point on the water level reading was wrong.

March 11, 21:30: IC reopened

At around this time the operator on duty recognized that the IC Valve 3A indicator lamp showing the green “closed” status was about to go off. The operator was concerned about the risk of being unable to restart the IC due to battery drain. In addition, by this time he had learned that the condenser cooling water would not run out with several hours of operation and reasoned that the small amount of steam from the “pig snout” was not due to the drop of water level in the condensate tank. Although the operator thought that probably the fail-safe function had closed the valves, he also thought there still was a chance of activating the IC and so turned the Valve 3A switch to open. Upon this action, sound of steam release was heard for a while, but then again it faded away. The operator believed that IC was not operating and reported that he had opened Valve 3A to the response headquarters. The response headquarters, not aware of the “close” operation at 18:25, could have found the report puzzling, but no one did.

March 11, 21:51: apparent rise in the radiation dosage

A rise in the dosage led to an order to ban entry into the reactor building.

March 11, 22:00: wrong water level information

Water level measurement gave a value of TAF + 550 mm. This report was a sign of worsening misreading of rising water level despite no reason for the water level to rise.

March 11, about 22:30: large increase in the radiation dosage

An operator on duty tried to enter the reactor building, but in front of the double door, his pocket dosimeter indicated 0.8 mSv within about 10 s, and sensing the risk, he returned to the central control room. This reading was equivalent to as high as about 300 mSv an hour. For information, the standard safe exposure for a person throughout his life is about one-third of this value at 100 mSv.

March 11, 23:50: IC problem finally recognized

The staff measured the D/W pressure using a portable generator from a contractor in the central control room. The reading was extremely high at 0.60 MPa. Plant Manager Yoshida, on hearing this report, finally recognized that the IC was not properly working and that the steam that had leaked into the CV D/W from the RPV had produced the abnormal rise in D/W pressure.

The reason for this D/W pressure rise could have been caused by hydrogen and rare gas, which do not condense to water, discharged into the S/C from the RPV, and with other gas then entered the D/W; however, there was also the possibility of damage on the pressure vessel to let gas escape into the CV.

Plant Manager Yoshida, upon receiving the report of D/W pressure at an abnormally high value of 0.6 MPa, realized that the IC was not working. With the possibility of even higher pressure, without hesitation, he ordered the power generation team and recovery team of the response headquarters to prepare Unit 1 for venting. At the same time, he ordered to prepare Unit 2 for venting as well in case of worst-case scenario.

March 12, around 01:30: all parties agreed to the venting

At the main office response headquarters, President Shimizu had agreed to the venting of Units 1 and 2. TEPCO also requested approval for the operation from the Prime Minister’s Office and NISA and soon received approval from Prime Minister Kan and others.

March 12, 01:48: D/DFP stoppage and evaluation of the alternative fire engine

D/DFP of Unit 1 was running from the day before in preparation for its use; however, it was found to have stopped by this time. The reason for the stoppage is still unknown. The response headquarters reasoned that the only alternative left was to use a fire engine to inject water and started to evaluate that procedure. Water injection to the reactor with a fire engine was a procedure never evaluated before and, thus, faced a number of difficulties. First, none of the TEPCO employees knew how to operate a fire engine. TEPCO had to request a contractor on-site to perform the work, which was outside their contract. The response headquarters did not know where the inlet for injecting water into the reactor was, and extra time was needed to locate a contractor employee with such knowledge. Furthermore, the only nearby water supply was the fire reservoir, which had insufficient water for extended water injection.

March 12, about 02:30: sudden rise of D/W pressure

The D/W pressure was reported to be at 0.84 MPa (abs). Most persons involved would have thought that such a high pressure could cause explosive damage to the CV.

March 12, 02:45: sudden RPV pressure drop

At about this time, the RPV pressure was measured at 0.8 MPa. That was a large drop from 6.9 MPa, measured about 7 h earlier at 20:07 on the previous day (Figure 2.17). No depressurization was made with SRVs during this time, and without doubt, there was a large leakage from the pressure vessel (see Remarks). The rise in the D/W pressure to about the same 0.84 MPa backs this conjecture.

March 12, 03:06: press conference about the venting

The Minister of Economy, Trade and Industry, Banri Kaieda, and a TEPCO Executive, Akio Komori, held a press conference to announce the plan to vent. Up to this point, the response headquarters had erroneous optimistic information about the situation with Unit 1 due to misunderstanding the IC status and misreading of the water level monitor. On the other hand, until about after 02:00 on March 12th, because of not knowing the RCIC operation and having no water level reading, officials had thought that Unit 2 was the reactor in most danger. Immediately before this press conference, the government received the news that Unit 2 had its RCIC running, and its water level and pressure were stable. Executive Komori took the stand in the press conference without hearing the news, and when asked, “Which is the first unit to vent?” he gave a confusing reply.

March 12, after 04:00: freshwater injection started

Started to inject freshwater using a fire engine. The water was first supplied from the fire engine tank and then collected from the fire reservoir and poured into the fire engine tank. This intermittent injection amounted to only 1-2 ton every several tens of minutes. At this time, as shown in Figure 2.18, the D/W pressure was going down. Because D/W cooling (D/W spray), S/C cooling (S/C spray), nor venting were not made, the only reason for this pressure drop was leakage from the CV.

March 12, about 04:23: rise in radiation dosage

The radiation dosage at the monitoring post near the main gate (Main Gate MP) rapidly rose to 0.59 μSv/h and to about 1.8 μSv/h at around 05:20 (the reading was around normal at about 0.069 μSv/h at 04:00). Also the central control room for Units 1 and 2 was located on the second floor of the control building between the two units. The dose reading there was high at around 05:00. In addition, the dose measurement was higher where closer to Unit 1. The operators moved toward the Unit 2 side and waited. It was a sign of serious core damage in the Unit 1 reactor located at an elevation higher than the central control room.

March 12, 06:50: concern about not completing the vent

The Minister of Economy, Trade and Industry Kaieda ordered a vent in accordance with regulations. Around this time, the authorities in Tokyo expressed concern about the progress of the vent, and some even speculated that the operators were hesitating in the task. The reality was that a number of difficulties occurred in performing the vent.

March 12, 07:11: Prime Minister Kan’s visit to the site

Prime Minister Kan and NSC Chair Haruki Madarame arrived at the scene. Vice President of TEPCO Sakae Mutoh and Plant Manager Yoshida alone from the plant greeted them and explained that they intended to perform the vent at around 09:00.

March 12, 9:15: preparation for opening the vent line

At about this time, a worker went into the second floor of the reactor building and opened the motor-driven vent valve by about 25%. This work, as we described earlier, was necessary to open the vent line.

The radiation level on the second floor of the reactor building at this time was not so high, and a worker could reach there to manually open the valve by turning its handle (Figure 2.19). Then, however, the other air-operated (A/O) vent valve needed compressed air, and insufficient air further delayed the process. In all, it took another 5 h to succeed with the vent.

March 12, about 09:24: opening of the vent line further delayed

At about this time, an operator headed toward the basement of the reactor building to manually open the A/O valve (small). The basement floor at the time, however, had a high radiation level, and the worker had to retreat to the central control room.

March 12, about 10:17: misconception of vent success

The operator gave up the manual operation and tried to remotely open it from the central control room. He hoped that compressed air in the instrument air system (IA, see Remarks) piping might open the vent valve (A/O small valve). After this trial at about 10:40, the response headquarters received a report of radiation increase at the main gate and monitoring post and thought that the vent had succeeded; however, at around 11:15, the radiation level went down again, and headquarters had to correct this earlier conclusion.

March 12, about 12:00: decision made to switch to seawater

At about this time, Plant Manager Yoshida decided to switch to seawater in case the freshwater was used up, and he ordered the building of a seawater injection line with hoses. TEPCO employees and contractors searched for accessible seawater near the buildings. They soon spotted a large amount of seawater in the reverse flow cleaning valve pit (see Remarks; Figure 2.20).

At about this time, TEPCO staff, including VP Mutoh and NSC chief Madarame understood that the seawater injection process was the only choice and had no opposition to it.

March 12, about 12:30: vent preparation completed

Around this time, the workers finally obtained a portable air compressor from a contractor. It was needed to open A/O valves. This compressor, however, had a small capacity and had to be set close to the A/O valve and in an area with relatively low radiation. After evaluation, they decided to set it near the reactor building large equipment hatch. The operation, however, took efforts for not having prepared the adaptor and other reasons, but around 14:00, the connection was complete, and the compressor was ready to supply air. The process of opening the A/O valve required opening the solenoid valve on the air piping as well. The central control room, therefore, after starting up the portable compressor, applied AC current from the small generator for the temporary lamps to energize the solenoid.

March 12, 14:50: venting judged to have succeeded

With all the aforementioned work, the D/W pressure dropped to 0.58 from 0.75 MPa, and the image on the TV showed white smoke coming from the exhaust tower. Plant Manager Yoshida judged that the “vent was successful back at about 14:30” and forwarded the information to Ministries and Agents at 15:18.

March 12, 14:53: preparing for seawater injection

At about this time, the freshwater in the fire reservoir was used up, and there was no alternative source of freshwater at hand then. Plant Manager Yoshida thus ordered to inject seawater, and the line construction was almost complete at about 15:30.

March 12, 15:36: hydrogen explosion

Unit 1 had a hydrogen explosion, and five workers in the field were injured (Figure 2.21). For a while, the effect of the explosion had to be evaluated, and until safety was confirmed, the recovery work had to be put on hold. The new seawater injection line was damaged and could not be used.

The three fire engines luckily were found operable. The explosion, however, forced the team to restart the electricity recovery work, which had almost been complete, from scratch. Plant Manager Yoshida was greatly disappointed at the setback caused the hydrogen explosion. The workers at the time had to take cover in the Seismic Isolated Building.

March 12, about 17:20: discussion about the effects of seawater injection

After checking the damage, the field restarted work for seawater injection. At about the same time in the Prime Minister’s Office, Prime Minister Kan, Assistant Goshi Hosono, NSC Chief Haruki Madarame, Assistant Manager Eiji Hiraoka, and TEPCO Fellow Ichiro Takekuro had gathered in the prime minister’s office. When Prime Minister Kan asked about the effect of seawater injection to the reactor, both Chief Madarame and Fellow Takekuro answered, “Even if it is seawater, we have to put priority on injecting water as soon as possible.” Prime Minister Kan then asked Chief Madarame, “Will injecting seawater cause re-criticality?” and Chief Madarame answered, “You do not have to worry so much about a possible re-criticality.” However, Kan was not satisfied with the answer.

March 12, 19:04: seawater injection started

The seawater injection, whose preparation had been once blocked by the hydrogen explosion, finally started at 19:04.

Fellow Takekuro, who was not aware of the start, strongly requested Plant Manager Yoshida to “Halt the seawater injection because they are discussing it in the prime minister’s office now.” Yoshida was worried about worsening the situation by stopping the seawater injection and consulted with the TEPCO main office and VP Mutoh at the off-site center. They gave him the same reply: “We have to halt the injection because we do not have the blessing from Prime Minister Kan.” Yoshida then judged, however, that he should not stop the seawater injection and whispered to those on duty in a voice low enough to escape detection by speakerphone, “By whatever means, never stop the water injection,” and then, in a loud voice that reached all corners of the response headquarters room, he ordered, “Stop the seawater injection.”

Later, at about 19:30, discussion resumed in the prime minister’s office. Prime Minister Kan, unaware of the process, immediately approved the seawater injection, and Fellow Takekuro lost his chance to explain the situation to Prime Minister Kan. At 20:20, the order to restart seawater injection was given, and the water injection up to that point was recorded as “water injection test” to agree with the time of offical approval.

March 13, past 08:00: radiation increase

For about an hour after this time, radiation levels around monitoring post 4 (MP-4) went up rapidly. The rate, which was 35 μSv/h at about 8:01, rose to 1204 μSv/h at 8:33 (Figure 2.22). This value had 2 more digits compared to the data around the main gate just a few hours earlier.

The reason for this rise was probably not due to the Unit 1 hydrogen explosion or its venting. Most likely it was due to radiation leakage from the CV after core damage of Unit 1, Unit 3, or maybe both.

Unit 1 is estimated to have suffered damage to its CV by about 22:00 on March 11th, and the damage grew by the dawn of March 12th; however, the damage could have expanded more by this time.

March 14, about 01:10: seawater injection halted

At about this time, water withdrawal from the Unit 3 reverse flow cleaning valve pit was disabled, and seawater injection to Unit 1 stopped. The workers then started to build a seawater supply line using a newly arrived fire engine to withdraw seawater directly from the ocean and pour it in the reverse flow cleaning valve pit. The line was completed shortly after 09:00; however, injection to Unit 3 had priority then. To make things worse, Unit 3 had its hydrogen explosion 2 h later, and the fire engine stopped its operation. Water injection to Unit 1, thus, did not restart until about 20:30, meaning that injection to Unit 1 had stopped for over 19 h.

March 14, about 06:00: further radioactivity leakage

Data from CAMS was available by this time. The data showed higher γ-ray values in the D/W than in the S/C from this time on to late afternoon of March 15th. High γ-ray measurement inside the CV despite the already started leakage from the CV means that core damage was still in progress, and levels higher in the D/W than in the S/C were an indication of a high possibility of leakage from the RPV directly to the D/W (Figure 2.23).

Summary of damage to the reactor primary equipment

Table 2.2 shows the damage to the core, RPV, and CV of Unit 1, as they were confirmed by the Investigation Committee. The expressions are based on evidence and have been selected carefully. The locations of leakages from the RPV and CV have not been identified.

TEPCO’s check of the actual IC isolation valves

(1) On April 1, 2011, TEPCO investigated the IC MO valve circuits. If the fail-safe function “failed” to work and the valves were left as the operator intended, valve 3B should be at “closed” and the others at “open.” The operators did not turn the valves to the “closed” side on or after March 11. The investigation revealed that the electrical circuits of 1A and 4A of System A and 1B and 4B of System B were all indicating “partially open.” These indications were probably caused by the fail-safe function sending out “close” signals, but with the driving power to realize the “close” operation lost at the same time, the closing motions had stopped midway.

(2) TEPCO further, on October 18, 2011, checked the actual openings of the four valves outside the CV. The investigation found that valve 2B was closed, verifying that the fail-safe function worked. Reports say that after the SBO, valves 2A and 3A, which the operator hit open after the fail-safe function, were actually confirmed open.

Cooling water left in the condensate tank

On October 18, TEPCO checked the remaining water in the IC condensate storage tank. The tank for System A was 65% full and that for B at 85%. Normally, the amount of cooling water to store is about 80%, and compared to that level, System A was 15% less and System B at the initial state. The IC operation duration to consume 15% of cooling water is, from calculation, about 1 h, and this number backs the assumption that System B was not used, and System A ran intermittently for about 50 min before the tsunami attack and later hardly functioned at all.

Loss of natural convection due to hydrogen

The IC takes in high-temperature steam, cools and condenses it in the condenser, and recycles the water back into the RPV. The driving force, similar to convection, is the difference of specific weights of steam and water. When hydrogen gas lighter than steam mixes in the system, the lightweighted hydrogen stagnates in the upper part of the piping to stop the circulation. IC is a cooling system; however, it stops functioning when core damage takes place and hydrogen is produced. It has a problem in countering severe accidents.

All AC power was lost due to water submergence caused by the tsunami. DC power survived. The central control room shared with Unit 4 was pitch dark. Unit 3 had its DC power switchboard and batteries on the concourse floor (between the first floor and basement); thus, this equipment did not receive water damage and allowed the operators to monitor the RPV pressure, water level, and other crucial parameters. RCIC had started at 15:25 with the “reactor water level high” signal at 15:25 before the loss of AC power.

March 11, 16:03: RCIC manually started

At about this time, the operator on duty manually started the RCIC (remotely). Because DC power had survived in Unit 3, the central control room was able to operate the RCIC and other equipment as well as read parameters off the indicators.

The RCIC of Units 2 and 3 each had a rated capacity of injecting 97 ton of water per hour. The capacity of the condensate storage tank was about 2500 ton; thus, running the RCIC at full capacity would drain the condensate storage tank in about a day. Water in the condensate storage tank had other tasks and could not spare all its water for the RCIC. The operator had to control the RCIC so it could last through the long hours to come.

March, 12, 11:36: RCIC automatically stopped

The RCIC stopped for some reason, still unknown as of today. The operators looked for the cause and tried to restart it from the central control room, but did not succeed.

March 12, 12:35: HPCI automatically started

The HPCI automatically started on detecting the lowered water level in the reactor (HPCI, like RCIC, used the condensate storage tank or S/C as its source of water). HPCI is also driven by steam from the reactor and thus does not require AC power for running its pump. HPCI is the last resort of emergency core cooling that can inject water into the RPV, even at high pressure. Its rated capacity is 965 ton/h. This high performance quickly raises the reactor water level and stops automatically. Such repetition of ON and OFF will quickly exhaust the battery. The operator, worried about draining the battery, built a bypass route so part of the water discharged from HPCI would run back to the condensate storage tank to control the amount of water injected to the RPV.

The HPCI water injected effectively cooled the reactor, and its pressure gradually dropped. This pressure decrease lowered the HPCI pump discharge pressure. At this time, however, the reactor water level had not been read, and the effect of water injection was not known. This uncertainty caused concerns in the operator’s mind about malfunction of the HPCI, which was forced to run in a state different from normal operation.

March 12, 17:30: early preparation of vent line ordered

The D/W pressure at this time was at about 0.3 MPa (abs), about twice the normal pressure. Plant Manager Yoshida, with the experience of taking long hours to vent Unit 1, felt the need to ready the vent line quickly before the radiation level rose in the reactor building and gave orders to hurry in preparing for the vent.

By that time, the plant emergency headquarters, having learned from the experience with Unit 1, were looking into the procedure for venting Units 2 and 3. The investigation revealed that the two units, unlike Unit 1, did not have handles on the large and small A/O valves for venting the RPV. In other words, to perform the vent, compressed air was absolutely necessary. The M/O valves, on the other hand, had handles for manual operation. It seems it is a common practice for nuclear power plants to put manual handles on M/O valves.

At 02:42, the operator manually stopped the HPCI. As described earlier, the operator on duty had concerns about the status of HPCI and, worried if he should stop it, had consulted with the plant response headquarters. Some of the engineers in the headquarters discussed the matter and reached the conclusion that stopping the HPCI could not be helped for the following reasons.

(1) Running HPCI at a rotation slower than its design speed may lead to damage to the HPCI.

(2) Opening the SRVs would lower the reactor pressure to allow low-pressure water injection using the alternative route from D/DFP.

The team relayed their conclusion to the operator; however, this discussion stayed within the team with the fact that HPCI was stopped. The information of HPCI stopping was not shared with the team leader or Plant Manager Yoshida. As we will discuss later, this process may have affected the later progression of events. When the operator was about to manually turn the HPCI off, the SRV indicator was illuminated green, indicating the SRVs were stopped. The fact that the lamp was on meant that the battery was still alive. The operator judged that he could operate the SRVs from the central control room. After stopping the HPCI, however, when the operator tried to activate the SRVs twice at 02:45 and 02:55, they would not open. It is probable that the battery capacity had dropped to a level to illuminate the status indicator lamps, but below a level to “open” the SRVs, a process that requires a large amount of current. This conjecture is based on the success of their openings, 7 h later, when the batteries had recovered.

March 13, about 03:00: alternative water injection from D/DFP was unsuccessful

The HPCI had successfully cooled the reactor to drop its pressure to 0.58 MPa. After stopping, however, the pressure quickly rose to 0.77 MPa at 03:00 on March 13th, and to 4.1 MPa at 03:44 (Figure 2.26). The D/DFP, which is a low-pressure system, did not have enough discharge pressure to inject water into the reactor vessel at a higher pressure. The situation was critical because they depended on water injection from D/DFP and stopped the HPCI, but the reactor pressure increased to a high level that did not allow water injection from D/DFP. The operators that lost methods of injecting water tried to restart HPCI or RCIC in vain. (The failure to restart the HPCI was probably also due to the low battery.)

Plant Manager Yoshida and other main staff at the plant response headquarters learned about this HPCI stoppage shortly before 04:00, more than an hour after it took place. Yoshida believed before hearing this report that alternative water injection from D/DFP was unreliable because even its closest water source of “filtered water tank” for Unit 1 was located over 500 m away in a straight line, and the piping could have been damaged by the earthquake (the piping seismic classification was level C), and the discharge pressure of about 0.4 MPa was lower than that of the fire engines. Yoshida had a chance of turning down the idea of manually stopping HPCI if he had been consulted about the process by those on-duty before they actually carried out the process.

Shutting down the HPCI should have been done after verifying an alternative method for water injection was available (i.e., opening the SRVs and testing water injection from D/DFP).

March 13, 04:50: serious preparation for venting

At about this time, the rising reactor pressure without means for water injection forced workers to energize the solenoid valves to open the A/O valves for venting. The operators brought in a portable generator for temporary lighting into the central control room and hooked the cables to the solenoid valve terminals trying to activate the valves. Later, however, when an operator went into the basement of the reactor building to check the A/O valve status, it was closed. The operator that entered the torus room heard the sound of SRV discharging steam into the S/C. When he put his foot on the S/C, it was so hot it partly melted the sole of his shoe.

The operator then, at around 05:23, changed the air bottle to one with air remaining in it and carried out the work to open the large A/O valve. After this work, the recovery team from the plant response headquarters headed to the basement of the reactor building to check the opening of the A/O valve. However, by this time, the radiation level in the area was too high to allow entry, and they had to go back to the central control room. They were still unsure if the A/O valve had actually opened or not.

March 13, about 05:00: information from water level indicator became available

At about this time, for an unknown reason, the reactor water level data became available. The water level indicator showed − 2300 mm from the TAF at about 05:10 and − 2600 at about 06:00. From the dropping water level, the plant response headquarters assumed it had dropped to TAF at about 04:15. The water level indication at this time was fairly reliable.

At around 07:39, D/W spray (an emergency cooling system that cools inside the D/W with water spray) was activated with water from D/DFP to lower the D/W pressure. At this time the D/DFP system was available.

March 13, about 06:00: preparation of seawater injection from fire engines

From the previous night of March 12th to about this time, seawater injection to Unit 1 was made by connecting three fire engines in series and pumping seawater up from the Unit 3 reverse flow cleaning valve pit. The sudden stoppage of the Unit 3 HPCI forced the workers to start alternative water injection to Unit 3 as well. After 06:00, the other fire engine on standby moved out to form a seawater injection line for Unit 3. The preparation was complete around 07:00.

About this time, however, on the fifth floor of the prime minister’s office, Minister Kaieda, Chief Madarame, and others were discussing matters like “Injecting seawater will result in decommissioning the reactor” or “Maybe there is freshwater remaining somewhere” and they decided to inquire if there was freshwater anywhere in the plant. A department manager of TEPCO informed Plant Manager Yoshida about the discussion, and Yoshida acknowledged it as a “strong opinion from the prime minister’s office to inject freshwater.” He then reconfigured the water injection line from the completed seawater line to a freshwater line from the fire reservoir (Figure 2.27).

The freshwater injection started almost 7 h after stopping the HPCI at about 09:25. The water in the fire reservoir only lasted for about 3 h, and the injection stopped at about 12:20. Then the workers had to reconfigure the water injection system back to seawater injection, and it had to wait until about 13:12 to start. When we look at the loss of water injection, the freshwater injection not starting until 09:25 was not due to switching the water source, but it was the time lost in waiting for the reactor pressure to drop to the level to allow the injection. The 52-min loss, however, for switching back to seawater after exhausting all the freshwater was due to the change made earlier to freshwater. This loss was an example of directives from Tokyo delaying the team in the field.

March 13, 07:44: securing 12 V batteries

The batteries delivered to the plant from outside up to this point were all 2 V batteries, which were inadequate for activating the SRVs that required 120 V. The plant response headquarters started to search for 12 V batteries and by 07:44 secured 10 batteries from the response headquarters employee cars parked at the plant. These batteries were carried into the central control room in preparation for SRV opening. At around 9:50, over 7 h from stopping the HPCI, the SRVs were opened to release pressure; however, as we will explain later, the pressure vessel had already been depressurized before that from damage to it.

March 13, 08:41: vent line complete

At 08:35, an operator on duty entered the second floor of the reactor building to turn the handle on the vent M/O valve to 15% open. The other A/O valve had already been opened about 4 h earlier. The state was unconfirmed because they could not enter the building. However, at 08:41, the plant response headquarters judged that the vent line was ready (i.e., automatic venting would take place when the CV pressure rises above the rupture disk set point). The headquarters reported so to the ministries and agencies.

March 13, about 09:00: possible damage to the RPV

At about 09:00 on March 13th, the reactor water level shot up, reactor pressure dropped almost instantly, and the D/W pressure rose rapidly (Figure 2.28). The D/W pressure then indicated values that were about the same with the RPV pressure or higher at 0.647 MPa. These data suggest the possibility that part of the fuel dropped to the bottom of the RPV and upon contacting the cooling water at the bottom, generated a large amount of steam to boost the pressure. The RPV bottom was damaged to cause a rapid pressure drop (RPV melt-through). The water level data that followed recorded an abrupt fluctuation. At this time the water level indication may have been out of order, but at the same time, could be showing the real transition of water level jumping up and down with the melted fuel dropping to the RPV bottom and contacting the water to produce an exploding amount of steam.

The Interim Investigation Report (December 2011) states that this rapid pressure drop was from opening the SRVs; however, from the hearings that followed, the committee found it was not. The Final Report corrected the error, and SRV depressurization only took place at about 09:50.

March 13, 09:20: the vent

The D/W pressure around 09:10 showed a peak value of 0.637 MPa (gauge) above the rupture disk set point of 0.427 MPa (gauge), but 14 min later it dropped to 0.540 MPa. The plant response headquarters determined the S/C was vented at around 09:20 and reported so to the ministries and agents. The two vent valves had been open since about an hour earlier at 08:41, and the venting system was ready to vent automatically as soon as the pressure rose above the rupture disk set point. Further at about 10:55, the D/W pressure dropped to 0.27 MPa (Figure 2.29). This suggests a large amount of gas exhaust with radioactive material was made around this time from the exhaust stack to the atmosphere.

March 13, 11:17: D/W pressure rose again, difficulty with keeping the vent valves “open”

Around this time, the D/W pressure started to rise again. The plant response headquarters estimated that pressure drop in the air bottle caused the opening of the vent valves to narrow. Workers entered the reactor building with high radiation levels in two shifts to take turns in 15-min intervals to exchange the bottles. At 12:30, the large A/O valve was confirmed “open,” and the D/W pressure, which was at 0.48 MPa at 12:40 dropped to 0.23 MPa at 14:30.

The D/W pressure, however, soon turned around at about 15:00 and went up to reach 0.41 MPa at about 16:45. The plant response headquarters decided there was a limit to being able to keep changing the bottles by entering the building with high radiation and started to discuss a new method using a compressor. The headquarters then managed to borrow a portable compressor from a contractor and started it at around 19:00. The compressor, however, was a small one, and the D/W pressure was still at 0.425 MPa at 20:30. After about 2 h at around 22:30, probably because all the piping was filled with air, the D/W pressure started on its way down to 0.24 MPa at 0:00 on March 14th. The headquarters for the first time managed to keep the large A/O valve open. So from the time they first opened the vent valve at 8:41 on March 13th, they had struggled with it for over 15 h.

Now, let’s roll back the time by about half a day.

March 13, 13:12: seawater injection started again

As we showed earlier, freshwater in the fire reservoir was exhausted, and after 52 min of interruption, the water source was switched to seawater in the reverse flow cleaning valve pit, and the water injection started again.

March 13, about 14:31: a number of dangerous symptoms

At about this time, a high level of 300 mSv/h was measured at the north side of the double door, and a white haze was seen inside the door. The workers feared a hydrogen explosion like that in Unit 1, but there was no effective way of stopping it. They even considered shooting a hole open on the reactor building wall with a fighter jet from the self-defense force. They had to abandon the idea because of the possibility of igniting the hydrogen and finally settled on using a water jet. They later started to prepare the tools; however, before they were ready, the hydrogen exploded the next day.

The radiation level in the central control room went up to 12 mSv/h in the Unit 3 side, and the operators had to take cover in the Unit 4 side.

March 14, 01:10: seawater in the reverse flow cleaning valve pit was exhausted, and water injection from fire engine stopped

At about 01:10, the seawater in the reverse flow cleaning valve pit was exhausted, and the fire engine that was injecting water into Unit 3 could no longer draw seawater. The plant response headquarters had been searching for a new water source since the day before (e.g., the water intake near Unit 4, maintenance access hatch over the water discharge channel, directly pumping water up using the fire engine, the basement of the turbine building, and many other possibilities); however, they all were unfeasible. When the fire engine stopped drawing water, another observation of the reverse flow cleaning valve pit revealed there was still some water left and, from about 03:20 on March 14th, managed to restart the water injection. That was probably just a short-term solution, and the headquarters, instead of relying on seawater left in the pit, started to discuss directly pumping water up from the ocean. The primary buildings were standing on land 10 m above sea level; thus, pumping the water with one fire engine would not work, but if they connected two in series, the seawater could be pumped. The workers started building the line, and after 09:00, they completed a line that continuously pumped seawater from the ocean into the reverse flow cleaning valve pit.

March 14, about 03:40: energizing the small A/O vent valve

As Figure 2.29 shows, the vent valve that seemed stabilized went unstable, and at about 03:40 the D/W pressure started again on its way up. Headquarters decided to open the bypass small valve in addition to the main large A/O valve. The operators used the portable small generator for temporary lighting to energize the solenoid of the instrument air (IA) system. Then at about 05:00, they switched the compressor to the portable one delivered from Fukushima-2. The D/W pressure, nonetheless, kept rising until about 07:00 and peaked at 0.52 MPa. Thus, all these efforts did not have as much effect as hoped for.

The D/W pressure transition on March 13th and 14th shown in Figure 2.29 indicate how the D/W pressure continued going up and down because the A/O valve opening was not stable.

March 14, 06:30: workers took cover

Around this time, the D/W pressure was staggered at above 0.5 MPa in a manner similar to Unit 1 before its hydrogen explosion, and workers were concerned with hydrogen leakage from the CV and hydrogen explosion to follow. Plant Manager Yoshida was concerned with the workers’ safety,and after consulting with the main office, he ordered temporary retraction to the seismic isolated building around 06:30-06:45. The D/W pressure, however, did not rise any higher, and because building the line to pump seawater from the ocean had priority, Yoshida lifted the order after 07:30.

March 14, 10:53: self-defense force water truck arrived

Around this time, the self-defense force arrived with seven water trucks carrying 35 tons of freshwater. They were headed toward the reverse flow cleaning valve pit in front of the turbine building of Unit 3. When two of the trucks approached the pit, Unit 3 had its hydrogen explosion. Four men on the self-defense force were injured, and the water truck suffered damages. No water was supplied from this truck then.

March 14, 11:01: Unit 3 hydrogen explosion

After a little over 3 h from the lifting of the retraction order, at 11:01, Unit 3 had the hydrogen explosion that everyone was concerned about (Figure 2.30). This explosion injured four personnel from the self-defense force, four TEPCO employees, and three employees of a subcontractor (Nanmei). The workers in the field all took cover in the seismic isolated building. Rumors of a large number of employees missing created a heavy mood in the plant response headquarters. The four fire engines parked in front of the turbine building all stopped working, and the fire hoses were damaged and unusable. The area was covered with rubbish from the explosion.

Reactor parameters of Unit 3, however, were still available after the explosion. The D/W pressure was at 0.48 MPa and S/C at 0.47 MPa; thus, Yoshida judged that it was a hydrogen explosion of the reactor building, similar to the one with Unit 1.

March 14, after 13:00: response restarted

Observation of the situation in the field revealed that many of the fire engines had stopped with their hoses unusable from the damage. The method of injecting water up to the point before the hydrogen explosion was to pump seawater from the ocean to the reverse flow cleaning valve pit using two fire engines in series and with different fire engines to inject the seawater individually to Units 1 to 3. The area around the Unit 3 reverse flow cleaning valve pit was covered with rubble, and it seemed difficult to rebuild the lines. The headquarters decided to directly inject the seawater from the ocean to Units 2 and 3. During this time, with priority to carry out water injection to these two units, water injection to Unit 1 was halted.

March 14, about 16:30: water injection restarted

The water injection line damaged with the hydrogen explosion was recovered and water injection to the Unit 3 reactor restarted.

March 15, 7:55

Steam was observed above the building, and discussion was made if they should first spray water on the Unit 3 fuel pool instead of the spent fuel pool of Unit 4 with the most number of fuel bundles. Later the spray was first sent onto Unit 3 (Section 2.9).

Later

RPV pressure and D/W pressure gradually went down. The CV pressure drop was not rapid; thus, CV damage with Unit 3 was not as bad as that with Unit 2.

Summary of damage to the reactor primary equipment

Table 2.4 shows the damage to the core, RPV, and CV of Unit 3, as they were confirmed by the Investigation Committee.

Water submergence from the tsunami caused SBO including DC. The central control room was pitch dark, and all the indicators were not functional. The most important water level and pressure indications of the reactor were unavailable as well. Fortunately, the RCIC had been manually (remotely from the central control room) turned on, but its status from that point was unknown.

March 11, 20:49: temporary lighting turned on, status unknown

A temporary light was turned on in the shared central control room for Units 1 and 2. The water level and injection status, however, were still unknown, and the response headquarters, as preparation against worse situations, prepared to request evacuating the area near the plant.

March 11, about 21:15: pessimistic conjecture of the status

The plant response headquarters, not knowing the RCIC status, estimated the water level to drop to TAF at 21:40, core damage to start at 22:20, and RPV damage at about 23:50. Headquarters assumed the worst for not knowing the water level and other parameters (contrary to their assumption, core cooling was in fact maintained).

March 11, 22:00: water level found to be in a stable condition

Water level was found at TAF + 3400 mm. Around this time, the RPV temperature was at near normal operation level, and the water level indication was also reliable. After about an hour and a half, the RPV pressure was at 6.3 MPa (gauge) and D/W pressure at 0.141 MPa (abs), both in normal ranges, and the engineers were relieved for the moment.

March 12, 01:00-02:55: RCIC confirmed to be operational

At past 01:00, an operator headed toward the RCIC room in the basement of the reactor building to check its status. When he opened the door, however, water gushed out and not being able to enter the room, he headed back to the central control room. At past 02:00, the operator went down to the RCIC room again and managed to get to the entrance, but could not check the RCIC status. He then came up with the idea that he could check the RCIC status by comparing the RPV pressure and the RCIC discharge pressure. He then went to the instrumentation racks on the first and second floors of the reactor building and found the RPV pressure at 5.6 MPa (gauge) and RCIC discharge pressure at 6.0 MPa (gauge). That is, he confirmed the RCIC discharge pressure was slightly higher than the RPV pressure, values that show that the RCIC was running normal. The head of operators on duty then judged that RCIC was operational and reported so to the plant response headquarters.

March 12 about 04:30: RCIC water source switched

The condensate storage tank for each unit was located in front of the turbine building. The capacity was 2500 tons for Units 2 and 3. The operator, concerned with the risk in keeping the RCIC running with the condensate tank, switched the water source to the S/C from the condensate storage tank.

March 12, 17:30: order to prepare the vent

From the experience of delaying the vent of Unit 1, Yoshida ordered an early vent preparation. Venting required compressed air to open the A/O valve; however, the only portable compressor available was in use for Unit 1. They decided to open the A/O valve (Figure 2.31) with the air from the air bottle on the IA system piping.

March 13, 10:15: vent preparation

With the plant manager order at 17:30, an operator went into the reactor building and opened the M/O valve at “25% open.” In addition at around this time, the plant manager had ordered to open all valves on the vent line except the rupture disk (i.e., to open the A/O valve as well). The recovery team of the plant response headquarters went into the first floor of the reactor building and opened the discharge valve on the air bottle next to the IA system piping to open the A/O valve. They further used the portable generator for temporary lighting to force the solenoid on the large A/O valve open by energizing it. At about 11:00, they succeeded in opening the large A/O valve to complete the vent line preparation. The S/C pressure at this time, however, was smaller than the rupture disk set pressure of 0.427 MPa (gauge), and the venting did not take place then.

March 13, past 18:00: preparing a compressor

The plant response headquarters decided to stop using air bottles with unknown amount of air left and rely on a portable compressor like Units 1 and 3. Through teleconference, they requested one from other plants, and about 7 h later, one arrived from Fukushima-2 before 02:00 on March 14th (Unit 1 had one since about 12:30 on March 12th and Unit 3 from about 19:00 on March 13th from contractors).

March 13, late in the afternoon: water injection line ready for Unit 2

At about this time, a water injection line was readied for Unit 2 like Units 1 and 3 that used the reverse flow cleaning valve pit as the water source. Unit 2, at this time, however, had its RCIC in operation and the RPV pressure was still a little less than 6 MPa. Water injection with a fire engine was only made ready and kept on standby.

March 14, 11:01: Unit 3 hydrogen explosion, vent line and water injection lines damaged

Unit 3 had a hydrogen explosion. By that time, the central control room had prepared a temporary circuit to energize the A/O valve solenoid. The hydrogen explosion, however, broke the circuit, and the large A/O valve reclosed. The explosion also damaged a fire engine and hoses, but luckily, the compressor survived the blast.

As described earlier, the reactor water level was going down after around 12:00. The water source for the RCIC had been switched to the S/C pool water from around 04:00 on March 12th, meaning the cooling water had been storing heat for over two days. Thus, the S/C at about 12:30 was at an abnormal state of high temperature and pressure with temperature at about 149 °C and pressure at 0.49 MPa. Given the situation, Plant Manager Yoshida judged that “the RCIC had stopped at about 13:25.”

March 14, 14:43: delay in seawater injection

At about this time, the water injection line damaged by the hydrogen explosion had been recovered early in the morning, thanks to great efforts by the workers. The afterquakes, however, had interrupted further work, and the fire engine started only at about 16:30, and water injection to Unit 3 started then. Unit 2, however, took several hours in operating the SRVs delaying the reactor depressurization, and the state of no water injection continued.

March 14, about 15:00: D/W pressure dropped

The D/W pressure at 0.465 MPa (abs) at about 12:30 started to drop after 13:35, and at about 16:57, it had clearly dropped to 0.40 MPa (Figure 2.32). During this time, no venting or CV cooling system activation had taken place. The D/W pressure drop was, thus, probably due to leakage from the CV that had been subject to high temperature and pressure for an extended time.

March 14, about 16:00: vent valve did not open

The recovery team of the plant response headquarters kept energizing the solenoid from the central control room; however, as mentioned earlier, the team could not keep the A/O valve open. This was possibly due to the small size of the compressor (see Remarks). At 18:35, the operators tried to open the small valve, in addition to the large A/O valve. They succeeded in cracking it open; however, at this time, the CV pressure had not reached the rupture disk set pressure and in any case, the status of no venting continued.

March 14, about 16:34: SRVs did not open

At about this time, the operators in the central control room connected the ten 12 V batteries collected from around the plant in series and tried to activate the solenoids of the SRVs; however, the SRVs did not open.

March 14, 18:22: all fuel bundles exposed

At 18:22, the water level gauge indicated TAF − 3700 mm, and at 18:50 it went off scale at the lower end, and the water level could no longer be measured. The plant and main office response headquarters both recognized that all fuel bundles were exposed from water at about 18:22.

March 14, 19:03: RPV depressurization with SRVs

At about this time, the SRVs finally opened, and the reactor pressure went down to 0.63 MPa to allow water injection. At about this time, Unit 3 had seawater injection drawn directly from the ocean and relayed with two fire engines connected in series with hoses. Unit-2 was ready for water injection in a similar manner without having to pour the seawater into the reverse flow cleaning valve pit.

March 14, about 19:20: the fire engines ran out of gas

At about this time, water injection to Unit 2 had just started; however, the fire engines were found to have run out of gas. TEPCO’s self-fire defense team, in response, supplied gas from their gas tank trucks; however, water injection to Units 2 and 3 was suspended for at least 37 min.

March 14, 19:57: continuous water injection to Unit-2 started but unstable

Unit 2 without any water injection after its RCIC stoppage, finally received continuous water injection from a fire engine.

Although the water injection had restarted, the Unit 2 RPV pressure was most of the time (over half of the time) from about 21:00 on March 14th to about 01:00 the next day, at a level higher than 1 MPa. Thus, during this time, water injection from the fire engine was physically impossible. Damage to the core and RPV progressed then, and the D/W pressure probably went up and stayed there.

March 14, about 23:35: decision made to vent the D/W

The biggest problem with Unit 2 at this time was that it had not succeeded in depressurizing the CV with S/C venting. The RPV had the safety valve SRVs, whereas the CV did not have equivalent equipment. If the vent continued to be unavailable, the “last resort” CV had the danger of exploding.

At about this time, the S/C pressure was showing values lower than the rupture disk set point, and the D/W, on the other hand, was showing higher values of 0.7 MPas above the rupture set point (see Remarks). The staff at the response headquarters, including the main office, reluctantly decided to vent the D/W. As mentioned earlier, D/W venting was different from S/C venting because it would release the gas inside the CV directly into the atmosphere without passing it through water and thus would cause radiation release that was more than one hundred times that from S/C venting. It was a process to avoid at all costs; however, it was judged that protecting the CV from destruction had higher priority.

March 15 about 00:16-01:11: highly dangerous state with high CV pressure

As Figure 2.33 shows, the RPV pressure exceeded 1 MPa three times between 20:30 on March 14th and 01:30 on March 15th. These large fluctuations can be explained as follows: With the RPV already damaged and leaking, the melted fuel and water came into contact to cause strong evaporation of the water to steam that rapidly increased the RPV pressure, and then with leakage from the RPV, the pressure dropped again. The phenomena repeated to produce the cycle in Figure 2.33.

Next as Figure 2.32 shows, the D/W pressure went down to about 0.4 MPa from about 14:00 to about 17:00 on March 14th. In general, when a vessel has leakage, its pressure will keep going down; however, as Figure 2.34 shows, the D/W pressure quickly went up around 23:00. The reason for this transition is as follows: The CV had already started its leakage, and after about 20:00, large leakage from the RPV started as well. Thus the CV was leaking to the outside but at the same time had RPV leaking into it. The rapid rise in the D/W pressure around 23:00 was caused by stronger leakage into the D/W from the RPV.

The radiation (γ-ray) monitoring results from CAMS (Figure 2.35) also support this conjecture. At 15:15 on March 14th, S/C was at about 0.01 Sv/h and D/W its tenth at about 0.002 Sv/h. After 6 h and 40 min of interruption, the restarted measurement at 21:55 showed that the D/W radiation had rapidly risen to exceed that of the S/C (Figure 2.34). This was probably due to leakage from the RPV directly to the D/W. Thus, the rapid and large D/W pressure rise starting at about 23:00 in Figure 2.34 matches the data from CAMS.

Summarizing the situations around this time, water injection from the fire engine to the RPV with pressure above 1 MPa was impossible. The CV venting failed, and the D/W pressure rapidly went up around 23:00 on the 14th to reach 0.7 MPa in about 30 min and stayed at this dangerous level, which could have resulted in CV explosion. In the midst of such a dangerous situation, Plant Manager Yoshida then consulted with the main office response headquarters and decided to keep, in the worst-case situation, only necessary personnel for each plant control and to evacuate the others from the Fukushima-1 site. To avoid a sense of panic to the others, he gave instructions to only a few to prepare buses so workers could flee the site if the situation worsened.

There is no question that both the RPV and CV had suffered damage at midnight on March 14th; however, the staff did not think that leakage from the CV had started. They were concerned with not being able to vent the CV and were worried about the worst-case situation of its explosion. A loud sound was then heard at 06:10.

March 15, 06:00-06:10: loud sound

It was when the shift was about to turn over when they heard a loud explosive sound in and around the central control room. They were all then given instructions to take cover in the seismic isolated building, and when they exited the building, they were surprised to see that the surrounding scenery had drastically changed. This exploding sound, as we will discuss later, was in fact the sound of the hydrogen explosion of Unit 4. The situation that had surrounded the staff at the time of not succeeding with the vent and the D/W pressure stopped at a high value of 0.7 MPa had put the staff in a state where everyone was holding his breath. The big bang shook the building, and many of them feared the worst. Under the situation, the plant response headquarters kept the minimum necessary staff of 50 on site and evacuated the remaining 650 workers temporarily to Fukushima-2.

The S/C pressure immediately after the explosion was pointing at zero, and with the sound they heard, the people thought that the CV had exploded. The time that the S/C pressure dropped to zero was, however, a few minutes before the exploding sound. Also, this S/C pressure indication was in absolute pressure, and an absolute pressure of zero would mean vacuum, and that physically could not have happened. For these two reasons, the judgment that “the CV exploded and its pressure dropped to zero” was wrong. Also, as Figure 2.34 shows, the S/C pressure was pointing at about half the D/W pressure and suddenly dropped to zero at 06:02. A situation that the S/C pressure was half of that of the D/W pressure for an extended time is also physically hard to imagine (see Remarks). From these facts, we can reason that the S/C pressure indication was unreliable and the values were wrong.

After about an hour, at 07:20, the D/W pressure recorded 0.73 MPa (abs) and then after another long interruption of over 4 h at 11:25, it went down to 0.155 MPa, a value close to atmospheric pressure. It is probable that the CV suffered significant damage during this time.

Later, the monitoring post near the main gate showed extremely high radiation values on March 15th and 16th with a peak at 10,000 μSv/h (around 08:00 on March 15th). The radiation was the highest on March 15th to 16th, and unfortunately, the wind changed into a relatively strong southeast one blowing from the ocean toward the land (Figure 2.36). Also at night, 1.5 mm of cold rain fell for over an hour. This wind and rain dropped a large amount of radiation over the area northwest of the plant and caused serious radiation contamination. These were unfortunate weather conditions for the local districts of Iitate, Namie, and others northeast of the plant.

March 15, 11:25: lifting the temporary evacuation order

As mentioned earlier, the high D/W pressure of above 0.7 MPa was found to have dropped to 0.155 MPa. Those that had evacuated Fukushima-1 to take cover in Fukushima-2 gradually returned to the plant response headquarters starting with group managers.

Later

The RPV and D/W pressure values remained at about the same level with the atmosphere. Radiation reading from CAMS for the D/W did not drop for a long time.

Summary of damage to the reactor primary equipment

Table 2.5 shows the damage to the core, RPV, and CV of Unit 2, as they were confirmed by the Investigation Committee.

The tremor from the explosion was caught on a number of seismographs, and the time of the sound was 06:12. The difference between the arrival time of the primary wave (P wave) and shear wave (S wave) at multiple locations allow spotting the source. The data identified the source to be around Unit 4 and not Unit 2. The exploding sound heard at 06:12 was, without question, the hydrogen explosion in Unit 4, and because there was no evidence of another explosive sound, there is no reason to believe that the Unit 2 CV exploded.

About “zero S/C pressure”

When the workers heard the exploding sound, the S/C pressure went to zero, and many thought that there was an explosive destruction around the S/C. TEPCO’s data, however, recorded S/C pressure at zero twice at 06:02 and 06:10 before the hydrogen explosion in Unit 4 (06:12). And as we mentioned earlier, the S/C pressure indication was the absolute pressure, and there is no explanation to support that the S/C was at vacuum. Also, the D/W pressure around this time was at 0.7 MPa, and from the CV structure, the two pressure values of the D/W and S/C were unlikely to have large differences. The S/C pressure gauge of Unit 2, thus, was likely broken at this time. Data from pressure gauges in general are reliable. This was the only event with a troubled pressure gauge in the Investigation Report.

Therefore, the thoughts of many workers in the field that there had been a big explosion with the CV near the S/C and the S/C pressure drop to zero resulted from the fear of such an event in the people’s minds. The hydrogen explosion in Unit 4 happened to take place at about the time of zero S/C pressure indication.

On the other hand, there is hardly any question about a large-scale leakage of radioactivity from Unit 2 from March 15th on, and the denial of an explosion does not mean the damage to the Unit 2 CV was small.

The only explosive material was hydrogen

Both Units 1 and 3 maintained RPV and CV pressure values several times that of atmospheric pressure; thus the possibility of hydrogen explosion with the RPV or CV is denied. If such an explosion took place inside the vessels and blasted the building walls, the vessels would have been destroyed as well, and their pressure values would have been at the atmospheric level.

Also, time had passed since the earthquake and thus there was no chance of dust from the shaking of the earthquake causing a dust explosion. An explosion from combustible gas was the most likely cause, and because there was no combustible gas in an amount to possibly cause an explosion other than hydrogen, it was identified as the cause.

There was enough hydrogen to cause an explosion

To have a hydrogen blast on the fifth floor of Unit 1, assuming the volumetric low limit of hydrogen explosion at 18.3%, about 312 kg of hydrogen was needed. On the other hand, TEPCO analysis announced in March 2012 and the Japan Nuclear Energy Safety Organization (JNES) analysis announced in September of the same year both concluded that about 900 kg of hydrogen had generated in the reactor. Thus the core damage had probably produced enough hydrogen to blow up the fifth floor of Unit 1.

For Unit 3, similar calculation returns 371 kg as the necessary amount of hydrogen. TEPCO’s analysis of the generated hydrogen was over 600 kg and that of JNES 550-700 kg. Even if the vent had released some into the air, there was plenty of hydrogen left to explode.

The hydrogen explosion of Unit 4 took place on the fourth floor, which had about a fifth of the volume compared to its fifth floor. The minimum amount of hydrogen necessary to cause an explosion was calculated to be 74.2 kg, and this amount is reasonable as an amount that trickled in from Unit 3 (Figure 2.41).

Reasons why Unit 2 didn’t explode

When the fifth floor of the reactor building of the adjacent Unit 1 exploded, the blowout panel of Unit 2 opened like a window, and hydrogen never accumulated in the top floor (Figure 2.42). What seemed like damage from an event turned around and saved the building.

We often hear that this blowout panel not popping open with Units 1 and 3 was a backfire from a lesson from the trouble at Kashiwazaki-Kariwa NPP during The Chuetsu Offshore Earthquake. During this 2007 earthquake, a blowout panel dropped and created some concerns. Later as a countermeasure, involving other plants as well, the panel was secured tighter to its frame. The Investigation Committee, however, learned that the modification had not been in place yet for Unit 3. The blowout panel, nonetheless, did not pop open even with the high CV pressure.

After the hydrogen explosion in Unit 1, the headquarters saw that Unit 3 would also have a hydrogen explosion for sure, and discussions were begun to open a hole in the reactor building. The conclusion was to use water jet; however, the plan was not completed in time, and the explosion took place on March 14th.

The principle of nuclear reactor water level sensing is to measure the water pressure change at the bottom of the pressure vessel to detect the cooling water level (Figure 2.54). If there was a water level change of 2 m, the change in pressure then is about 0.02 MPa.

If the water level sensor simply measures water pressure at the bottom of the pressure vessel as shown in Figure 2.54a, the measured pressure is the cooling water pressure (assume the water level at about 10 m) 0.1 MPa plus the steam pressure of about 7 MPa to bring the reading to about 7.1 MPa. Then the measurement range is too large for accurately measuring changes around the size of 0.02 MPa.

The sketch in Figure 2.54b solves this problem by measuring the difference between the steam pressure at the top of the pressure vessel and the water pressure at the bottom of the reactor. This difference shows the reactor water level at accuracy several tens of times better.

The water level sensors in actual nuclear reactors are as Figure 2.54c shows. The reference condensate pot at the top of the reference side of the pressure gauge allows measuring the difference between the reactor water level and the water level in the reference condensate pot. This configuration further improves the accuracy of water level measurement.

This configuration, however, as we will discuss later, has a fundamental problem that if the water level in the reference condensate pot goes down with the RPV temperature rise, the water level reading goes wrong. Some countermeasures against this problem should be made for the case of severe accidents.