MOORE, Circuit Judge.

On February 15, 1951 the M/S Christer Salen, one day out of Yokohama, bound for Vancouver, B.C., was proceeding at a speed of approximately 12 knots into a heavy sea, when without warning and for no apparent reason she suddenly split in two at about the middle of No. 3 hold.

The forward part, after being jammed by a sea into the after part, floated off and sank with the consequent loss of all cargo therein. The after part returned without incident to Yokohama. The cargo in No. 4 and No. 5 holds sustained only slight damage.

The cargo owners filed the libels in which the interlocutory decrees here appealed from were entered. The vessel owners filed cross-libels against the owners of the saved cargo, claiming contribution for general average, and petitioned for a limitation of liability.

Subsequent tests revealed that the break was "a brittle fracture" originating in the starboard sheerstrake, i.e., the uppermost longitudinal hull plating on the starboard side, somewhere in its 7½" extension above the main deck.

The M/S Christer Salen, a general dry cargo ship, 405 feet long, 56 feet wide, with a gross tonnage of 4,900 tons, was one of a class of nine ships built at the Eriksberg Yard, Gothenburg, Sweden. The steel for the ship was rolled in Germany in 1943 and her construction was completed in 1944. Her hull was all welded and riveted to the frames. She began operating in 1945 at which time she obtained the standard ship classification of Lloyd's Register of Shipping, namely, 100-A-1. She retained this classification until she split in two.

In January 1950 she ran aground in the Philippines; permanent repairs were made in the Swedish yard, and in May 1950 a certificate of seaworthiness was issued by Lloyd's, continuing her 100-A-1 classification. In November-December 1950 while the Christer Salen was en route from Vancouver, B.C. to Kobe heavy weather caused a crack at the top of its portside bulwark welded to the port sheerstrake where the bulwark joined the bridge front. This crack was v-ed out and welded. Between January 6 and January 8, 1951, only a month before the disaster, the ship was drydocked at Hong Kong for normal maintenance purposes, viz., cleaning, scraping, and painting, and inspected by a Lloyd's surveyor who found that her "shell plating" was "in good condition" and recommended continuance of its classification (Exh. NN).

Were it not for the forceful arguments advanced by the appellant cargo owners, we would have thought it a foregone conclusion that no liability could attach to the carrier under the Carriage of Goods by Sea Act, 46 U.S.C.A. section 1304(2)(p) and (q), which provides that "Neither the carrier nor the ship shall be responsible for loss or damage arising or resulting from * * * (p) Latent defects not discoverable by due diligence; and (q) Any other cause arising without the actual fault and privity of the carrier and without the fault or neglect of the agents or servants of the carrier * * *." To understand the appellants' argument a brief description of the physical phenomena producing the fracture is necessary, which phenomena were explained in great detail at the trial by experts called by the parties.

In the 1940s disasters similar to that which befell the Christer Salen occurred in some other classes of cargo ships having all-welded hulls. Liberty ships seemed particularly prone to this affliction and in fact were the only class of dry cargo ships which displayed any such affinity. Some tankers having allwelded hulls, most notably T-2's, also had a tendency to fracture. This tendency to fracture was not and never has been present in ships having riveted hull seams.

The first explanation for major fractures not attributable to external forces, such as extremely heavy weather, occurring in welded ships was that they were monolithic structures, which owing to their lack of elasticity possessed locked-in stresses. On the other hand, on riveted ships there is a certain amount of play between the plates, and, while the plates may be sprung when subjected to sudden and extreme stress, the riveted seams will generally absorb the internal stresses which cause the major fractures on welded hulls. This explanation is now hotly disputed by protagonists of the welded hull. They claim that poor steel and poor design will produce as many fractures in riveted ships as they do in welded ones. What prevents major fractures in riveted ships is that the break will stop at the first riveted seam in place of running all the way down the hull.

The fact that a fracture will stop at a riveted seam prompted the superimposition of riveted bands, called crack arresters, on those welded ships, such as T-2's and Liberties, having a tendency to fracture. There are two varieties of crack arresters; in one type longitudinal slots are made in the hull or deck, with riveted straps then fitted over the slots; the other type consists of riveted gunwale bars inserted between the stringer plates and sheerstrakes. The evidence conclusively shows that crack arresters do the job which they were designed to do. The arguments advanced by the shipowner based on the testimony of Murray, a representative of Lloyd's, to justify Lloyd's failure to require crack arresters on the vessel were that the fitting of crack arresters might have caused notches which could have been the starting point of fractures and that crack arresters do not stop cracks.

The theory that crack arresters might have done more harm than good is pure speculation contrary to scientific facts and conclusively disproven by actual experience with crack arresters. Commenting upon the efficacy of crack arresters the report of the United States Navy Ship Structure Committee published in 1948 — three years before the accident here — stated that "The evidence that cleavage fractures do not pass riveted seams is now based on 15,000 ship-years service" (Exh. 21, p. 4). There is no evidence of any major fracture originating at crack arresters, and good workmanship would prevent the occurrence of any notches. Furthermore, in 1950 Lloyd's could not have been overly impressed by this fancied danger, since in 1947 it required crack arresters on sixty T-2's classified by it.

The claim that crack arresters are ineffective is also at odds with established fact. The sole support in the record for this position is the fact that two T-2 tankers fitted with crack arresters broke up in a severe storm off Cape Cod in February 1952. Even accepting as true the proposition that the cracks were not stopped at the crack arresters on these two ships, it is difficult to see how Lloyd's could rely on this fact in 1950 and 1951 since the accidents did not occur until 1952. But the proposition is not true. Lieutenant Commander MacCutcheon, one of the foremost authorities on structural failures in welded ships, had knowledge of the crack-ups of these two T-2's and testified that the fractures on these two vessels did not pass through the arresters, but actually stopped at the arresters and then ran back to them. The fact that crack arresters will not transform a ship into an impregnable floating fortress immune from all sea perils does not mean that they are ineffective.

There was testimony that the fractures which pose the gravest threat to the integrity of the ship are those which occur in notch brittle steel as opposed to those which occur in ductile steel. In notch brittle steel the break will be sharp, whereas in ductile steel the break will be fibrous. As temperature lowers, any steel will lose some of its ductility. Those which lose most of their ductility at normal operating temperatures are classified as notch brittle. Notch brittleness is characterized by a high carbon to manganese ratio. The steel in the Christer Salen's starboard sheerstrake had a high carbon content and became notch brittle at 30° F.

Even an all-welded hull of notch brittle steel which has no crack arresters and is subjected to heavy stresses, however, will not fracture usually unless a notch is present in one of the plates. Such a notch must therefore have been present in the extension of the starboard sheerstrake of the Christer Salen. The notch could have been put into the sheerstrake either by external damage, or by a latent defect in a weld or in the steel itself.

I. The Duty to Install Crack Arresters

Against this background appellants' arguments can be set forth and assessed. Appellants claim that since the Christer Salen had notch brittle steel at normal operating temperatures, and since crack arresters were well known as an effective means of preventing fractures on all-welded, notch brittle ships, the owner of the Christer Salen was under a duty to install crack arresters. They point out that this installation could have been easily accomplished when she was at her home yard in 1950 at the relatively insignificant expense of $10,000.

The ship owner, however, did not know that the Christer Salen was notch brittle and therefore susceptible to a brittle fracture, and it is our opinion that in the exercise of due diligence a reasonably prudent owner would not have been required to install crack arresters on the Christer Salen prior to 1951 because it was unforeseeable that the ship was actually notch brittle.

Appellants rely on the facts that the steel in the Christer Salen was rolled in Germany in 1943 and that after July 1943 Lloyd's became suspicious of the quality of German steel and thereafter would not accept the stamps of approval by Lloyd's agents in Germany but would conduct tests to verify the results on the German certificates. However, no test in 1943 or 1944 was in existence which would reveal the notch brittle quality of a steel, and at that time the dangers of such steel were just being uncovered. Liability cannot be imposed for failure to safeguard against a hazard not then generally known in the shipbuilding industry and for failure to conduct nonexistent tests.

Appellants' argument therefore comes down to the following proposition: that any owner of an all-welded ship who did not definitely know that his ship was not notch brittle was required at the earliest feasible time to outfit her with crack arresters. We think such a rule imposes too broad a standard of liability. The Christer Salen had been roaming the seven seas for years without displaying the slightest indication of her latent unseaworthiness; the operation of her eight sisterships likewise furnished no clues of any inherent danger.

Had the Christer Salen been of a class with a known predilection for cleavage fractures, such as Liberties, appellants would have a far stronger case. As it is, the Christer Salen did not contain any design features similar to those which were responsible for the vast majority of major fractures in Liberties. Over 75% of the fractures on Liberties started either at the sharp-edged hatch corners or at the cutouts made into the top of the sheerstrakes amidships for accommodation ladders. The hatch coamings on the Christer Salen were rounded and her gunwale tops were smooth.

Appellants also claim that Lloyd's requirement in 1947 that T-2's have crack arresters is some evidence that it should have done likewise with the Christer Salen. Aside from the fact that T-2's also displayed marked tendencies to fracture, the dissimilarity in design between tankers and dry cargo ships prohibits any such inference. The stresses in a tanker are generally much more severe than in a dry cargo vessel and are concentrated in the lower portion of the hull, whereas the stresses in a dry cargo ship are most severe in the uppermost part of the hull.

The recent case of States Steamship Co. v. United States, 9 Cir., 259 F.2d 458, is distinguishable for similar reasons. In that case the SS Pennsylvania sunk with all hands, apparently because of a major fracture. The twenty-two foot crack in her deck which occurred on the voyage immediately preceding the disaster should have put the owner on notice that he had an unseaworthy vessel.

II. Was the Notch a Discoverable Defect

As an alternative theory appellants claim that the notch was a discoverable defect, or at the very least the ship owner has not shown that it was not. The trial court rejected this theory in finding that the notch was a latent defect. We reach the same result as did the court below but through a different reasoning.

The basis for finding that the notch was a latent defect present in the sheerstrake when the ship was built was the expert testimony of Murray and Ibison, two ship surveyors for Lloyd's. While both men were emphatic in concluding that the notch was a latent defect, they were in disagreement on the nature of the notch. Ibison testified that it must have been a slag inclusion within the steel of the sheerstrake itself. Murray rejected this hypothesis and stated that it must have been a micro-crack in the center of a weld or in a weld junction.

In our opinion it is not necessary and virtually impossible to state with any degree of certainty what actually caused the fracture.

On this phase of the case, all the ship owner need establish is that normal precautions were taken to see that a discoverable notch did not exist. Appellants rely on the fact that there is no affirmative proof that such a defect was not present when the ship left Yokohama. However, if positive evidence were required to establish a negative proposition of fact (here absence of a discoverable defect), no carrier could ever discharge its statutory burden of proof. A month prior to the fracture a third party, Lloyd's ship surveyor at the Hong Kong drydock, inspected the shell and found it to be in good condition. This party was qualified to make the inspection and the inspection was made in the regular course of his business. Unless there is some evidence of bias, incompetence or fraud, the report (Exh. NN) must be accepted at its face value. Appellants have stipulated that the Hong Kong surveyor would have testified in accordance with this report (Exh. A).

From the time the inspection was made to the date the Christer Salen put out of Yokohama, she was not subjected to heavy weather. Neither the testimony of the master and chief mate nor the records kept by the ship reveals any accident which could have caused noticeable damage to the sheerstrake. The captain testified that he observed no damage to the sheerstrake and that no such damage was reported to him.

Appellants point out that the master was unaware of the potential danger in welded ships of cleavage fractures and therefore took no special pains to scrutinize the tops of the sheerstrakes. This lack of knowledge coupled with the failure to conduct a rigorous inspection, it is argued, precludes a finding of due diligence. While a master is presumed to know the general characteristics of his ship, he is not required to possess the combined knowledge and skill of a naval architect and metallurgist.

Upon all the evidence the trial court was clearly justified in finding that the accident was not attributable to a lack of due diligence on the part of the ship owner. The facts and law applicable thereto cannot support a conclusion that the ship owner should be held liable for this unfortunate but unforeseeable disaster.

The decree appealed from is affirmed.